betaFeedback

WorksPhysics and Astronomy

Oscillation modes, Decoherence, Beam Transfer Functions, and Instabilities for a Beam Subjected to Space-Charge Forces

Abstract

Based on the oscillation modes of a bunch subject to the space-charge force, we discuss the decoherence of the dipole oscillation, the beam transfer function, and the beam instability. The longitudinal phase space is discretized into nl×nJn_l\times n_J grids in the azimuthal and radial directions, respectively. The transverse dipole moment in the longitudinal phase space is represented by a vector of length nlnJn_l n_J, and the transformation given by the Vlasov equation is represented by a matrix transformation of size nlnJ×nlnJn_l n_J\times n_l n_J. Solving the eigenvalue problem of this matrix gives the nlnJn_l n_J eigen-oscillation modes of the bunch. Using the resulting eigenvectors within a standard linear-algebraic treatment, we analyze these three phenomena. The analysis was performed for the Rapid Cycling Synchrotron (RCS) of the China Spallation Neutron Source (CSNS) using nl=41n_l=41 and nJ=40n_J=40, and it visualizes the features related to the continuous distribution of the tune, Landau damping, and isolated modes separated from the continuous tune region.

Introduction

The behavior of beam particles moving in a circular accelerator can be described by equations of motion and/or a Hamiltonian. The primary objective is to determine x(s)x(s) at a given ss for an initial set of coordinates and momenta x(s=0)x(s = 0). We generally deal with the six-dimensional phase space represented by (x,px,y,py,z,δ)(x, p_x, y, p_y, z, \delta). Here, to handle coupled oscillations in the transverse xx and longitudinal zz directions, we use the four-dimensional subspace (x,px,z,δ)(x, p_x, z, \delta). The y−zy-z oscillation can be treated similarly to the x−zx-z oscillation.

To investigate the collective behavior of a beam, we consider the phase-space distribution of particles, Ψ(x)\Psi(\boldsymbol{x}). The motion of a single particle is described by equations of motion and a Hamiltonian, H(x)H(\boldsymbol{x}), which depend only on the phase-space coordinates of that particle; the effects of the other particles are incorporated by averaging. The equation of motion governing Ψ(x)\Psi(\boldsymbol{x}) is called the Vlasov equation and is expressed as follows:

∂Ψ∂s+∂Ψ∂x∂H∂px−∂Ψ∂px∂H∂x+∂Ψ∂z∂H∂δ−∂Ψ∂δ∂H∂z=0.(1)\frac{\partial\Psi}{\partial s}+\frac{\partial\Psi}{\partial x}\frac{\partial H}{\partial p_x}-\frac{\partial\Psi}{\partial p_x}\frac{\partial H}{\partial x}+\frac{\partial\Psi}{\partial z}\frac{\partial H}{\partial\delta}-\frac{\partial\Psi}{\partial\delta}\frac{\partial H}{\partial z}=0. \tag*{(1)}

The Vlasov equation can also be expressed using Poisson brackets:

∂Ψ∂s=[H,Ψ]=:H:Ψ.(2)\frac{\partial\Psi}{\partial s}=[H,\Psi]=:H:\Psi. \tag*{(2)}

Here, the Poisson bracket is expressed by the operator ::::. A formal solution can be written as:

Ψ(x;s)=exp⁡(:H:s)Ψ(x;0).(3)\Psi(\boldsymbol{x};s)=\exp(:H:s)\Psi(\boldsymbol{x};0). \tag*{(3)}

(4) has the opposite sign of HH compared to Hamiltonian equations of motion:

dxds=−[H,x].(4)\frac{d\boldsymbol{x}}{ds}=-[H,\boldsymbol{x}]. \tag*{(4)}

This is a difference in perspective: focusing on the particle distribution versus on the motion of individual particles. The distribution of a single particle can be represented using the Dirac delta function as

Ψ(x;0)=δ(x−xi),(5)\Psi(\boldsymbol{x};0)=\delta(\boldsymbol{x}-\boldsymbol{x}_i), \tag*{(5)}

and tracking the distribution using the Vlasov equation yields

Ψ(s)=δ(e:H(x):sx−xi)=δ(x−e−:H(xi):sxi).(6)\Psi(s)=\delta(e^{:H(\boldsymbol{x}):s}\boldsymbol{x}-\boldsymbol{x}_i)=\delta(\boldsymbol{x}-e^{-:H(\boldsymbol{x}_i):s}\boldsymbol{x}_i). \tag*{(6)}

It can be seen that the transformation of the individual particle coordinates xi\boldsymbol{x}_i follows Eq. (4), meaning that the distribution changes due to particle motion. In other words, the Vlasov equation states that the current phase-space density equals the past phase-space density, as the past density moves along with the particle motion. In Hamiltonian dynamics, the symplecticity ensures that the Jacobian of the transformation e:H:e^{:H:} is unity. The premise is that the phase-space density does not change with motion. The Vlasov equation is derived from a single-particle Hamiltonian for independent motion within an averaged field, and short-range two-particle collisions such as intrabeam scattering are not considered.

(3) is represented by a matrix transformation once the phase space is discretized. However, if one attempts to discretize the six-dimensional phase space, the matrix becomes extremely large. Therefore, by restricting the treatment to the dipole mode x(z,δ)x(z,\delta) in the xx direction alone, it suffices to discretize the longitudinal phase space. Here, for a beam subjected to the space-charge force, we obtain the eigen-oscillations and eigenvectors (eigen-oscillation distributions) by solving the eigenvalue problem for the transformation matrix of the dipole-mode distribution, represented by a vector obtained by discretizing the longitudinal phase space into

CircumferenceCC (m)227.9
Injection EnergyE0E_0 (GeV)0.08
Bunch populationNpN_p0.75×10130.75 \times10^{13}
Bunch lengthσz\sigma_z (m)25
Slippage factorηp\eta_p−0.81-0.81
Synchrotron tuneνs\nu_s0.01
Chromatic phaseχ/ξ\chi/\xi−1.2-1.2
Space charge tune shiftΔνsc\Delta\nu_{\mathrm{sc}}−0.27-0.27

Table 1. Parameters of CSNS used as a numerical application.

4141 (azimuthal) ×40\times40 (radial) =1640= 1640, i.e., a matrix of size 1640×16401640 \times1640.

Starting from the oscillation modes of a bunch undergoing synchro-betatron oscillations in an accelerator, we discuss the oscillation modes of a bunch oscillating under the influence of space-charge forces. As application examples, we discuss (i) the oscillations initialized by a uniform offset, (ii) the oscillations under sinusoidal forced excitation, i.e., the beam transfer function, and (iii) the beam instability that appear when an impedance is added.

Studies on the head-tail oscillation modes and instabilities have been carried out in many papers [1, 2, 3, 4, 5, 6]. They were based on the airbag square-well (ABS) model in the longitudinal phase space, which is easy to handle analytically. Numerical studies extending this approach have been carried out in Ref. [4]. Experiments and analyses for (i) kicker excitation and (ii) forced oscillation have been carried out based on the ABS model in Refs. [7] and [3, 8]. Coupled-bunch instabilities are analyzed based on the ABS model in Ref. [9]. The numerical treatment using the circulant matrix model is presented in Ref. [6].

The analysis presented here has been used to study coupled-bunch instability [10]. Here we focus on the 16401640 oscillation modes (eigenvectors) obtained by solving the eigenvalue problem of a 1640×16401640 \times1640 matrix. Our method allows us to handle arbitrary longitudinal distributions. The above subjects (i)-(iii) are discussed by a unified linear-algebraic technique; Landau damping is visualized in all three subjects.

This paper is organized as follows. The particle distribution for free synchro-betatron motion is analyzed in § 2: the basic formalism is presented in § 2 A, and the application examples (i) and (ii) are discussed in § 2 B and II C, respectively. The particle distribution with space-charge forces is analyzed in § 5: the basic formalism is presented in § 5 A and § 5 B, and the application examples (i), (ii), and (iii) are discussed in § 5 C, III D, and IV, respectively.

For a numerical analysis, the parameters of the China Spallation Neutron Source (CSNS) are used, as listed in Table I.

Synchro-betatron oscillations

Distribution of particles undergoing free synchro-betatron oscillations

The solution in the linear dynamics of a circular accelerator is well known. There are invariants in each degree of freedom, as

Jx=γxx2+2αxxpx+βxpx22,J=z2/βz+βzδ22.(7)J_x = \frac{\gamma_x x^2 + 2\alpha_x x p_x + \beta_x p_x^2}{2}, \qquad J = \frac{z^2/\beta_z + \beta_z\delta^2}{2}. \tag*{(7)}

For zz, we do not consider the tilt of the zz-δ\delta phase space; αz=0\alpha_z = 0. βz\beta_z is defined as the aspect ratio of the zz-δ\delta ellipse; βz=zmax⁡/δmax⁡\beta_z = z_{\max}/\delta_{\max} in the phase-space trajectories. Using the cavity parameters and the slippage factor (ηp\eta_p), it is expressed as βz=∣ηp∣Cβ2E0/(eV)/(2π)\beta_z = \sqrt{|\eta_p|C\beta^2E_0/(eV)/(2\pi)}, independent of ss, where CC is the circumference of a circular accelerator. For simplicity, we denote J=JzJ = J_z and ϕ=ϕz\phi= \phi_z for the synchrotron phase-space variables.

The Hamiltonian is expressed as

H(J;s)=Jxβx(s)−ηpJβz(8)H(J;s) = \frac{J_x}{\beta_x(s)} - \frac{\eta_p J}{\beta_z} \tag*{(8)}

or, considering the change in one revolution,

Hˉ(J)≡∮dsH(J;s)=2πνxJx+2πνzJ.(9)\bar{H}(J) \equiv\oint ds H(J;s) = 2\pi\nu_x J_x + 2\pi\nu_z J. \tag*{(9)}

The solution to the equation of motion shows that the betatron and synchrotron phases change in one revolution as Δϕx=2πνx=∮ds/βx(s)\Delta\phi_x = 2\pi\nu_x = \oint ds/\beta_x(s) and Δϕz=2πνz=ηpβzC\Delta\phi_z = 2\pi\nu_z = \eta_p\beta_z C.

The stationary solution is a distribution function Ψ(J)\Psi(J) that depends only on JJ.

Ψ(J,C)=exp⁡(Hrev(J))Ψ(J,0)=Ψ(J,0).\Psi(J,C) = \exp(H_{\mathrm{rev}}(J))\Psi(J,0) = \Psi(J,0).

The Poisson bracket of functions of the same variable vanishes, {J:ψ(J)}=0\{J:\psi(J)\} = 0.

The betatron variables are replaced by

xβx⇒xβxpx+αxxβx⇒px.(10)\frac{x}{\sqrt{\beta_x}} \Rightarrow x \qquad\frac{\beta_xp_x+\alpha_xx}{\sqrt{\beta_x}} \Rightarrow p_x. \tag*{(10)}

Thus, they are represented by Jx=(x2+px2)/2J_x = (x^2+p_x^2)/2, x=2Jxcos⁡ϕxx=\sqrt{2J_x}\cos\phi_x, px=−2Jxsin⁡ϕxp_x=-\sqrt{2J_x}\sin\phi_x. Further, we introduce v=x+ipxv=x+ip_x. With this, betatron oscillation as a function of ss is represented by a simple phase rotation:

v(s)=e−iϕxv(0),v(C)=e−2πiνxv(0).(11)v(s)=e^{-i\phi_x}v(0), \qquad v(C)=e^{-2\pi i\nu_x}v(0). \tag*{(11)}

While x(s)x(s) is used as a function of ss to describe motion, xx is also used as coordinates in phase space. Using the longitudinal coordinates (z,δ)(z,\delta) or (J,ϕ)(J,\phi) as phase-space coordinates, we treat the evolution of the averaged transverse phase-space variables,

v(z,δ;s)=∫(x+ipx)Ψ(x;s) dx dpxψ(z,δ)v(z,\delta;s)=\frac{\int(x+ip_x)\Psi(x;s)\,dx\,dp_x}{\psi(z,\delta)}
ρv(z;s)=∫(x+ipx)Ψ(x;s) dx dpx dδ(12)\rho_v(z;s)=\int(x+ip_x)\Psi(x;s)\,dx\,dp_x\,d\delta \tag*{(12)}
⟨v(s)⟩=∫ρv(z;s) dz,vˉ(z;s)=ρv(z;s)ρ(z;s)\langle v(s)\rangle=\int\rho_v(z;s)\,dz,\qquad\bar{v}(z;s)=\frac{\rho_v(z;s)}{\rho(z;s)}

where ψ(z,δ;s)=∫Ψ(x;s) dx dpx\psi(z,\delta;s)=\int\Psi(x;s)\,dx\,dp_x and ρ(z;s)=∫ψ(z,δ;s) dδ\rho(z;s)=\int\psi(z,\delta;s)\,d\delta are the longitudinal beam densities. ss is sometimes omitted when it is trivial.

Non-stationary oscillations encompass any distribution not expressible solely by JJ. Consider the case where the initial transverse distribution is given by f(Jx)f(J_x) in the xx-direction, with its center of mass in the longitudinal phase space given by v(J,ϕ;0)=x0(J,ϕ)v(J,\phi;0)=x_0(J,\phi), and the longitudinal distribution is ψ(J)\psi(J) around the center.

Ψ(x;0)=f(x−x0(J,ϕ),px)ψ(J)(13)\Psi(x;0)=f\left(x-x_0(J,\phi),p_x\right)\psi(J) \tag*{(13)}

in the xx-pxp_x and zz-δ\delta spaces. We can find ∫Ψ dx dpx=ψ(J)\int\Psi\,dx\,dp_x=\psi(J). The distribution Ψ\Psi is not an equilibrium distribution for x0≠0x_0\ne0, since it is not a function of JJ. As a typical example, in the case of a Gaussian in both planes, fGf_G and ψG\psi_G are

fG(x−x0(J,ϕ),px)=exp⁡(−(x−x0(J,ϕ))2−px22ϵx)2πϵx,(14)f_G\left(x-x_0(J,\phi),p_x\right)=\frac{\exp\left(-\frac{(x-x_0(J,\phi))^2-p_x^2}{2\epsilon_x}\right)}{2\pi\epsilon_x}, \tag*{(14)}
ψG(J)=e−J/ϵz2πϵz.(15)\psi_G(J)=\frac{e^{-J/\epsilon_z}}{2\pi\epsilon_z}. \tag*{(15)}

Integration over xx and pxp_x, after redefining the variable x−x0(J,ϕ)→xx-x_0(J,\phi)\to x, gives, as expected,

x(J,ϕ;0)=∫dx dpx x f(x−x0(J,ϕ),px)=x0(J,ϕ),(16)x(J,\phi;0)=\int dx\,dp_x\,x\,f\left(x-x_0(J,\phi),p_x\right)=x_0(J,\phi), \tag*{(16)}

where ff is an even function of the arguments x−x0x-x_0 and pxp_x; namely ∫dx dpx xf(x,px)=0\int dx\,dp_x\,x f(x,p_x)=0.

After one turn, the distribution function undergoes the inverse transformation e:H:e^{:H:} on xx, changing the betatron and synchrotron phases by −2πνx-2\pi\nu_x and −2πνz-2\pi\nu_z:

Ψ(x;C)=f(e2πiνxv−x0(J,ϕ−2πνz))ψ(J).(17)\Psi(x;C)=f\left(e^{2\pi i\nu_x}v-x_0(J,\phi-2\pi\nu_z)\right)\psi(J). \tag*{(17)}

By redefining the variables,

Ψ(x;C)=f(v−e−2πiνxx0(J,ϕ−2πνz))ψ(J).(18)\Psi(x;C)=f\left(v-e^{-2\pi i\nu_x}x_0(J,\phi-2\pi\nu_z)\right)\psi(J). \tag*{(18)}

Thus, the center-of-mass position in phase space after one revolution is

v(J,ϕ;C)=∫vf dx dpx=e−2πiνxx0(J,ϕ−2πνz).(19)v(J,\phi;C)=\int vf\,dx\,dp_x=e^{-2\pi i\nu_x}x_0(J,\phi-2\pi\nu_z). \tag*{(19)}

Here we introduce the synchro-betatron oscillation mode. Since vv is periodic with respect to the synchrotron phase, the azimuthal direction in phase space, we perform a Fourier expansion into azimuthal modes:

v(J,ϕ;s)=∑l=−∞∞vl(J;s)eilϕ.(20)v(J,\phi;s)=\sum_{l=-\infty}^{\infty}v_l(J;s)e^{il\phi}. \tag*{(20)}

Focusing on a mode with ll,

v(J,ϕ;C)=e−2πi(νx+lνz)x0(J,ϕ).v(J,\phi;C)=e^{-2\pi i(\nu_x+l\nu_z)}x_0(J,\phi).

The same argument holds if the initial offset is taken to be v0(J,ϕ)v_0(J,\phi) instead of x0(J,ϕ)x_0(J,\phi). Eq. (20) means that the azimuthal dependence of the distribution, v∼eilϕv\sim e^{il\phi} in the longitudinal phase space, is an eigenmode of the free synchro-betatron motion, with eigenfrequency (tune) νx+lνz\nu_x+l\nu_z.

Chromaticity

Chromaticity is a quantity characterizing the dependence of the betatron tune on the momentum deviation (δ=Δp/p0\delta=\Delta p/p_0) in single-particle motion. The tune is expanded with respect to the momentum deviation, and its coefficient is called the chromaticity. Here, only the first-order term is considered and is denoted by the linear chromaticity ξ\xi:

Δνx(δ)=ξδ.(21)\Delta\nu_x(\delta)=\xi\delta. \tag*{(21)}

In the case of a finite chromaticity, the ll-mode in the expansion of Eq. (20) is no longer an eigenmode. The ll-mode including the variation due to χ\chi becomes an eigenmode of oscillation, with tune νx+lνz\nu_x+l\nu_z [10]:

v(z,δ)=∑lvl(J)eilϕ+iχz/σz.(22)v(z,\delta)=\sum_l v_l(J)e^{il\phi+i\chi z/\sigma_z}. \tag*{(22)}

where the chromatic phase χ=2πξσz/(ηpC)\chi=2\pi\xi\sigma_z/(\eta_p C) is introduced. The variation due to chromaticity within the bunch is characterized by χ\chi. When the energy is below the transition, ηp\eta_p is negative, and the sign of χ\chi is opposite to the sign of the chromaticity ξ\xi.

Decoherence and recoherence of a kicked beam

Consider the case where the beam distribution is shifted by x0x_0 uniformly in the xx-direction as an initial condition. In practice, this is realized by giving an offset at injection or by kicking the entire bunch with a kicker magnet located π/2\pi/2 upstream in betatron phase.

Ψ(x)=f(x−x0,px)ψ(J).(23)\Psi(x)=f(x-x_0,p_x)\psi(J). \tag*{(23)}

For ξ=χ=0\xi=\chi=0, it is evident that the center of mass of the bunch undergoes betatron oscillation with tune νx\nu_x. The bunch oscillation then contains only the l=0l=0 mode. For a finite chromaticity χ≠0\chi\ne0, the eigenmodes with eigenfrequencies νx+lνz\nu_x+l\nu_z are eilϕ+iχz/σze^{il\phi+i\chi z/\sigma_z}, so a uniformly shifted distribution is represented by a superposition of multiple ll modes:

x0=∑lxl(J)eilϕ+iχz/σz.(24)x_0=\sum_l x_l(J)e^{il\phi+i\chi z/\sigma_z}. \tag*{(24)}

xl(J)x_l(J) is obtained by applying ∫dϕ e−ilϕ−iχz/σz\int d\phi\,e^{-il\phi-i\chi z/\sigma_z} to both sides:

xl(J)=x02π∫02πdϕ e−ilϕe−iχz(J,ϕ)/σz=x0i−lJl(χr),(25)\begin{aligned} x_l(J)=\frac{x_0}{2\pi}\int_0^{2\pi}d\phi\,e^{-il\phi}e^{-i\chi z(J,\phi)/\sigma_z} \\ &=x_0i^{-l}J_l(\chi r), \tag*{(25)} \end{aligned}

where r=2J/ϵzr=\sqrt{2J/\epsilon_z} can be viewed as a normalized radial coordinate in the longitudinal phase space. Jl(x)J_l(x) is the ll-th order Bessel function. The Bessel-function identity e−ixcos⁡ϕ=∑l=−∞∞i−lJl(x)eilϕe^{-ix\cos\phi}=\sum_{l=-\infty}^{\infty}i^{-l}J_l(x)e^{il\phi} ensures that summing over all ll yields the uniform shift x0x_0. In practice, ll is truncated at some value lmax⁡l_{\max}. Let us examine how many ll modes are needed to represent a uniform shift. Substituting Eq. (25) into Eq. (24) and truncating at ±lmax⁡\pm l_{\max}:

v(J,ϕ)=x0∑l=−lmax⁡lmax⁡i−lJl(χr)eilϕ+iχrcos⁡ϕ.(26)v(J,\phi)=x_0\sum_{l=-l_{\max}}^{l_{\max}}i^{-l}J_l(\chi r)e^{il\phi+i\chi r\cos\phi}. \tag*{(26)}

For zero chromaticity (χ=0\chi=0), the l=0l=0 mode alone can represent the uniform shift, as mentioned above. No dipole moment is associated with the oscillation of the l≠0l\ne0 modes. For χ≠0\chi\ne0, various ll modes mix, because they possess a dipole mode. Figure 1 shows the distribution for r<1r<1 in the case lmax⁡=8l_{\max}=8, χ=8\chi=8. It is distributed in 0.73<x(J,ϕ)<1.150.73<x(J,\phi)<1.15, roughly reproducing a uniform distribution. For χ=6\chi=6, 0.96<x(J,ϕ)<1.040.96<x(J,\phi)<1.04, and for χ=10\chi=10, 0.17<x(J,ϕ)<1.480.17<x(J,\phi)<1.48, so lmax⁡≥χl_{\max}\ge\chi is desirable. This makes sense when we consider that χ\chi is the wavenumber in the zz direction and ll is the wavenumber in the azimuthal direction. Corresponding to the shift in xx, the imaginary part of vv, px(J,ϕ)p_x(J,\phi), arises.

Figure 1. Reconstruction of intra-bunch phase amplitude for a uniform initial amplitude with lmax⁡=8l_{\max}=8, χ=8\chi=8.

Time evolution

Let us consider the time evolution of this uniformly shifted bunch. Each ll mode moves with its own frequency (νx+lνz)(\nu_x+l\nu_z):

v(J,ϕ;t)=2πx0∑li−lJl(χr)eilϕ+iχz/σze−2πi(νx+lνz)t.(27)v(J,\phi;t)=2\pi x_0\sum_l i^{-l}J_l(\chi r)e^{il\phi+i\chi z/\sigma_z}e^{-2\pi i(\nu_x+l\nu_z)t}. \tag*{(27)}

Here, time t=s/Ct=s/C is measured in turns. At t=0t=0, vv is real and only has an xx component, but an imaginary part pxp_x appears over time. Figure 2 shows the center of mass ⟨x⟩\langle x\rangle for 1000 turns. Even if the dipole amplitude disappears momentarily, it revives every 100 turns (1/νz1/\nu_z). Multiplying by the phase variation due to the tune in Eq. (12):

Oscillation of the bunch centroid $\langle x\rangle$ with initial amplitude 1, without space charge forces

Figure 2. Oscillation of the bunch centroid ⟨x⟩\langle x\rangle with initial amplitude 1, without space charge forces.

⟨x⟩(t)=x0∑l∫0∞Jl(χr)2ψ(r)r dr e−2πi(νx+lνz)t.(28)\langle x\rangle(t)=x_0\sum_l\int_0^\infty J_l(\chi r)^2\psi(r)r\,dr\,e^{-2\pi i(\nu_x+l\nu_z)t}. \tag*{(28)}

If the longitudinal distribution is Gaussian, ψ(r)=e−r2/2/(2π)\psi(r)=e^{-r^2/2}/(2\pi) (i.e., ψ(J)=e−J/εz/(2πεz)\psi(J)=e^{-J/\varepsilon_z}/(2\pi\varepsilon_z), the integral can be performed:

⟨x⟩(t)=e−χ2Re⁡[e−2πiνxt∑lIl(χ2)e−2πilνzt].(29)\langle x\rangle(t)=e^{-\chi^2}\operatorname{Re}\left[e^{-2\pi i\nu_x t}\sum_l I_l(\chi^2)e^{-2\pi i l\nu_z t}\right]. \tag*{(29)}

where Il(x)I_l(x) is the modified Bessel function. The periodic behavior with period 1/νz1/\nu_z can be inferred because the sum becomes large when tt is an integer multiple of 1/νz1/\nu_z.

Frequency response to forced oscillation: Beam Transfer Function

We set up a device that generates an electromagnetic field at high frequency at a specific location (s=0s=0) in the accelerator, and that gives a transverse momentum kick to the beam at a given frequency once per revolution:

Fx,ext(s,ω)=A∑ne−iω(s−z)/(cβ)δ(s−nC).(30)F_{x,\mathrm{ext}}(s,\omega)=A\sum_n e^{-i\omega(s-z)/(c\beta)}\delta(s-nC). \tag*{(30)}

A change of momentum dpx=F(0,ω) ds=Aeiωz/(cβ)dp_x=F(0,\omega)\,ds=Ae^{i\omega z/(c\beta)} occurs at s=0s=0. The corresponding Hamiltonian is H ds=−Aeiωz/(cβ)pxH\,ds=-Ae^{i\omega z/(c\beta)}p_x. The distribution function ff changes as follows:

e⋅H dsf(x,px)=f(x,px−Aeiωz/(cβ)).(31)e^{\cdot H\,ds}f(x,p_x)=f\left(x,p_x-Ae^{i\omega z/(c\beta)}\right). \tag*{(31)}

The average momentum px(z,δ)p_x(z,\delta) changes by:

px(z,δ;ds)=∫dx dpx pxf(x,px−Aeiωz/(cβ))=px(z,δ)+Aeiωz/(cβ).(32)p_x(z,\delta;ds)=\int dx\,dp_x\,p_x f\left(x,p_x-Ae^{i\omega z/(c\beta)}\right) =p_x(z,\delta)+Ae^{i\omega z/(c\beta)}. \tag*{(32)}

To find the ll component of Δpx=Aeiωz/(cβ)\Delta p_x=Ae^{i\omega z/(c\beta)}, we multiply both sides by e−iχz/σze−ilϕe^{-i\chi z/\sigma_z}e^{-il\phi} and integrate over ϕ\phi:

Δpl(J,ds)=A∫dϕ2πe−iχz/σze−ilϕ+iωz/(cβ)=AilJl((κ−χ)r).(33)\Delta p_l(J,ds)=A\int\frac{d\phi}{2\pi}e^{-i\chi z/\sigma_z}e^{-il\phi+i\omega z/(c\beta)} =A i^l J_l((\kappa-\chi)r). \tag*{(33)}

Here, the input frequency is normalized by κ=ωσz/(cβ)\kappa=\omega\sigma_z/(c\beta). Since HH includes zz, it also changes the longitudinal distribution ψ\psi, but this is ignored since it is small. Considering the repetition over ss, the change in vlv_l is described by a differential equation:

dvlds+2πCi(νx+lνz)vl=AilJl((κ−χ)r)e−2πinνδ(s−nC),(34)\frac{dv_l}{ds}+\frac{2\pi}{C}i(\nu_x+l\nu_z)v_l =A i^lJ_l((\kappa-\chi)r)e^{-2\pi i n\nu}\delta(s-nC), \tag*{(34)}

where ν=ω/ω0=ωC/(2πcβ)\nu=\omega/\omega_0=\omega C/(2\pi c\beta) is the tune of the input frequency. The solution of Eq. (34) is obtained by Fourier transformation, applying ∫ds e2πiν′s/C\int ds\,e^{2\pi i\nu' s/C} to both sides:

−2πiC(ν′−νx−lνz)∫ds e2πiν′s/Cvl(J;s)=Ail∫ds Jl((κ−χ)r)e2πi(ν′s/C−nν)δ(s−nC)=AilJl((κ−χ)r)∑pδ(ν−ν′−p).(35)\begin{aligned} -\frac{2\pi i}{C}(\nu'-\nu_x-l\nu_z)\int ds\,e^{2\pi i\nu' s/C}v_l(J;s) =A i^l\int ds\,J_l((\kappa-\chi)r)e^{2\pi i(\nu' s/C-n\nu)}\delta(s-nC) \\ &=A i^lJ_l((\kappa-\chi)r)\sum_p\delta(\nu-\nu'-p). \tag*{(35)} \end{aligned}

Here, the identity

∑n=−∞∞e2πinν=∑p=−∞∞δ(ν−p)\sum_{n=-\infty}^{\infty}e^{2\pi i n\nu}=\sum_{p=-\infty}^{\infty}\delta(\nu-p)

was used. By the inverse transform, applying ∫dν′ e−2πiν′s/C\int d\nu'\,e^{-2\pi i\nu's/C}, the amplitude vl(J)v_l(J) of the ll mode is obtained:

vl(J;t)=Ail+12π∑pJl((κ−χ)r)ν′−νx′−lνze−2πiνt.(36)v_l(J;t)=\frac{A i^{l+1}}{2\pi}\sum_p\frac{J_l((\kappa-\chi)r)}{\nu'-\nu'_x-l\nu_z}e^{-2\pi i\nu t}. \tag*{(36)}

where νx′\nu'_x is the fractional part of the tune νx\nu_x. In the sum over integer pp, ν′=ν−p\nu'=\nu-p is recast by adding the integer part of νx\nu_x: p+int⁡(νx)→pp+\operatorname{int}(\nu_x)\to p is used as the new pp. In this way, when the input frequency ν\nu is close to νx′+lνz\nu'_x+l\nu_z plus an integer pp, i.e., ν′∼νx′+lνz\nu'\sim\nu'_x+l\nu_z, the denominator in Eq. (36) becomes close to zero, forming a peak. It is common practice to perform tune measurements by applying a forced excitation at a frequency that differs from the beam betatron frequency by an integer multiple of the revolution frequency.

The condition for the response from modes excited by a sweep of the input frequency (ν)(\nu) to form a peak is that the denominator of Eq. (36) becomes zero: ν=νl,p≡p+νx′+lνz\nu=\nu_{l,p}\equiv p+\nu'_x+l\nu_z. Cases with ν<0\nu<0 are also considered. The tune (frequency) at which the response peaks is ν=νl,p\nu=\nu_{l,p}, corresponding to κl,p=ωl,pσz/(cβ)=2πνl,pσz/C\kappa_{l,p}=\omega_{l,p}\sigma_z/(c\beta)=2\pi\nu_{l,p}\sigma_z/C.

By calculating ρx(z)\rho_x(z) and ⟨x⟩\langle x\rangle for pp and ll, the magnitude of the response of each mode to the input frequency can

be evaluated. Let us focus on the numerator amplitude in Eq. (36), i.e., neglect AA, il+1i^{l+1}, etc., and obtain the relative amplitude for l,pl,p:

∣⟨vl,p⟩∣∼∫r drdϕ2πJl((κl,p−χ)r)eilϕ+iχz/σzψ(r)∼∫r drJl((κl,p−χ)r)Jl(χr)ψ(r).(37)\begin{aligned} |\langle v_{l,p}\rangle|\sim\int r\,dr\frac{d\phi}{2\pi}J_l((\kappa_{l,p}-\chi)r)e^{il\phi+i\chi z/\sigma_z}\psi(r) \\ &\sim\int r\,drJ_l((\kappa_{l,p}-\chi)r)J_l(\chi r)\psi(r). \tag*{(37)} \end{aligned}

In the case of χ=0\chi=0, the amplitudes for modes other than l=0l=0 are zero because of Jl≠0(0)=0J_{l\ne0}(0)=0. Figure 3 shows the amplitude ∣⟨vl,p⟩∣|\langle v_{l,p}\rangle| as a function of the fractional tune ν′\nu' of the input frequency for each chromaticity. The amplitude is plotted at the fractional input frequency ν′=νx′±lνz\nu'=\nu'_x\pm l\nu_z, and those for different pp are overlaid. A symmetric amplitude distribution appears around the tune νx=0.85\nu_x=0.85. At χ=0\chi=0, only the betatron mode ν′=0.85=νx′\nu'=0.85=\nu'_x responds strongly in several pp near p∼0p\sim0. At χ>0\chi>0, the maximum is at ν′=0.85\nu'=0.85, and the response becomes weaker as χ\chi increases. The relation between tune and amplitude for χ<0\chi<0 (ξ<0\xi<0) is similar to that for χ>0\chi>0 (ξ<0\xi<0). There are slight differences in how the fractional part of the tune is handled.

The dipole moment as a function of zz is also a measurable quantity,

ρx,l(z)∼∫dδ Jl((κl,p−χ)r)ψ(J)eilϕ+iχz/σz.(38)\rho_{x,l}(z)\sim\int d\delta\,J_l((\kappa_{l,p}-\chi)r)\psi(J)e^{il\phi+i\chi z/\sigma_z}. \tag*{(38)}

We are interested in the Fourier component of the dipole moment at the input frequency ω=ωl,p\omega=\omega_{l,p},

ρx(ωl,p)=∫dz ρx,l(z)e−iωl,pz/(cβ)∼∫r dr Jl((κl,p−χ)r)2ψ(r).(39)\begin{aligned} \rho_x(\omega_{l,p})=\int dz\,\rho_{x,l}(z)e^{-i\omega_{l,p}z/(c\beta)} \\ &\sim\int r\,dr\,J_l((\kappa_{l,p}-\chi)r)^2\psi(r). \tag*{(39)} \end{aligned}

Figure 4 shows the Fourier component ρx(ωl,p)\rho_x(\omega_{l,p}) at the input frequency. ρx(ωl,p)\rho_x(\omega_{l,p}) is plotted at the fractional input frequency ν′=νx′±lνz\nu'=\nu'_x\pm l\nu_z, and those for different pp are overlaid in plot (a). For χ=0\chi=0, ρx(ω0,p)\rho_x(\omega_{0,p}) with l=0l=0 has a peak value ≈1\approx1 at ω0,−1=(1−νx′)ω0\omega_{0,-1}=(1-\nu'_x)\omega_0. It has a decent value at p≤4p\le4. ρx(ωl,p)\rho_x(\omega_{l,p}) for l≥0l\ge0 is shown in plot (b). In Eq. (39), the integral can be evaluated using the integrand at r=1r=1: ρx(ωl,p)∼Jl(κl,p−χ)2\rho_x(\omega_{l,p}) \sim J_l(\kappa_{l,p}-\chi)^2. This behavior is consistent with plot (b). Note that Δκ/Δp=2πσz/C=0.689\Delta\kappa/\Delta p=2\pi\sigma_z/C=0.689.

Eq. (39) shows that the peak structure seen in plot (b) appears symmetrically for pp determined by κ0,p=χ\kappa_{0,p}=\chi. The peak structure in plot (b) shifts by Δp=C/(2πσz)Δχ=1.45Δχ\Delta p=C/(2\pi\sigma_z)\Delta\chi=1.45\Delta\chi as χ\chi is changed. Plot (c) shows the shift for χ=−4,0,4\chi=-4,0,4, which gives Δp=±5∼6\Delta p=\pm5\sim6 for Δχ=±4\Delta\chi=\pm4. These behaviors are reasonable, since the frequency matches the intra-bunch variation of chromaticity eiχz/σze^{i\chi z/\sigma_z} when κ=χ\kappa=\chi.

Oscillation of beams subject to space-charge forces

Formalism to treat space-charge force

Space-charge forces are forces acting between non-relativistically moving charged particles. Consider how space-charge forces affect oscillations within a bunch. Considering small oscillations of the bunch, a particle at x(z,δ)x(z,\delta) in phase space, or its distribution, experiences a space-charge force corresponding to the deviation from the center xˉ(z)=∫x(z,δ)ψ(J)dδ/ρ(z)\bar{x}(z)=\int x(z,\delta)\psi(J)d\delta/\rho(z) integrated over δ\delta at zz:

Δpx(z,δ)=kscβxρ(z)[x(z,δ)−xˉ(z)]Δs,(40)\Delta p_x(z,\delta)=k_{\mathrm{sc}}\beta_x\rho(z)[x(z,\delta)-\bar{x}(z)]\Delta s, \tag*{(40)}

where

ksc(s)=4Nrpβ2γ31Rx(Rx+Ry).(41)k_{\mathrm{sc}}(s)=\frac{4Nr_p}{\beta^2\gamma^3}\frac{1}{R_x(R_x+R_y)}. \tag*{(41)}

Using expressions WDW_D and WQW_Q for the dipolar and quadrupolar wake force integrated over the circumference,

WD(z)=kscCδ(z)WQ(z)=−WD(z),(42)W_D(z)=k_{\mathrm{sc}}C\delta(z)\qquad W_Q(z)=-W_D(z), \tag*{(42)}

the equation of motion for the space-charge force is expressed by

dpx(z,δ)ds=−βxC[∫WQ(z−z′)ρ(z′) dz′ x(z,δ)+∫WD(z−z′)x(z′)ρ(z′) dz′].(43)\frac{dp_x(z,\delta)}{ds}=-\frac{\beta_x}{C}\left[\int W_Q(z-z')\rho(z')\,dz'\,x(z,\delta)+\int W_D(z-z')x(z')\rho(z')\,dz'\right]. \tag*{(43)}

Impedance, which is Fourier transform of the wake force, is expressed as

ZD=kscCcβ≡ZscZQ=−Zsc.(44)Z_D=\frac{k_{\mathrm{sc}}C}{c\beta}\equiv Z_{\mathrm{sc}}\qquad Z_Q=-Z_{\mathrm{sc}}. \tag*{(44)}

The impedance is inherently a function of frequency ω\omega, but since the wake force is a delta function, it is a constant.

Here, we transform Eq. (40) into a form that converts a function of phase-space variables (z′,δ′)(z',\delta') into a function of (z,δ)(z,\delta):

dpx(z,δ)ds=−βxC[x(z,δ)∫dz′ dδ′WQ(z−z′)ψ(z′,δ′)+∫dz′ dδ′WD(z−z′)ψ(z′,δ′)x(z′,δ′)],(45)\frac{dp_x(z,\delta)}{ds}=-\frac{\beta_x}{C}\left[x(z,\delta)\int dz'\,d\delta'W_Q(z-z')\psi(z',\delta')+\int dz'\,d\delta'W_D(z-z')\psi(z',\delta')x(z',\delta')\right], \tag*{(45)}

As in Eq. (22), expand x,pxx,p_x into azimuthal ll modes. The equation of motion for each ll mode is obtained by multiplying by e−ilϕ−iχz/σze^{-il\phi-i\chi z/\sigma_z} and integrating.

dpl(J)ds=−βxC∫dϕ2π[∑l′xl′(J)e−i(l−l′)ϕ∫dJ′ dϕ′ WQ(z−z′)ψ(J′)+∫dJ′ dϕ′WD(z−z′)ψ(J′)xl′(J′)e−i(lϕ−l′ϕ′+ξ(z−z′)/σz)],(46)\frac{dp_l(J)}{ds}=-\frac{\beta_x}{C}\int\frac{d\phi}{2\pi}\left[\sum_{l'}x_{l'}(J)e^{-i(l-l')\phi}\int dJ'\,d\phi'\,W_Q(z-z')\psi(J')+\int dJ'\,d\phi'W_D(z-z')\psi(J')x_{l'}(J')e^{-i(l\phi-l'\phi'+\xi(z-z')/\sigma_z)}\right], \tag*{(46)}

The x(z′,δ′)x(z',\delta') on the right-hand side is also expanded with eil′ϕ′+iχz′/σze^{il'\phi'+i\chi z'/\sigma_z}. Using δ(z)=∫dω/(2πcβ)eiωz/(cβ)\delta(z)=\int d\omega/(2\pi c\beta)e^{i\omega z/(c\beta)} and the Bessel function formula:

dpl(J)ds=βxZscC∑l′il−l′xl′(J)[∫dωJl−l′(κr)D(κ)−∫dJ′ψ(J′)xl′(J′)∫dωJl(κr)Jl(κr′)].(47)\frac{dp_l(J)}{ds}=\frac{\beta_x Z_{\mathrm{sc}}}{C}\sum_{l'}i^{l-l'}x_{l'}(J)\left[\int d\omega J_{l-l'}(\kappa r)D(\kappa)-\int dJ'\psi(J')x_{l'}(J')\int d\omega J_l(\kappa r)J_l(\kappa r')\right]. \tag*{(47)}
Typeψ(J^)\psi(\hat{J})ρ(z)\rho(z)D(κ)D(\kappa)
G12πe−J^\frac{1}{2\pi}e^{-\hat{J}}12πσze−z2/2σz2\frac{1}{\sqrt{2\pi}\sigma_z}e^{-z^2/2\sigma_z^2}12πe−κ2/2\frac{1}{2\pi}e^{-\kappa^2/2}
W12πh(1−J^)\frac{1}{2\pi}h(1-\hat{J})1πσz22σz2−z2\frac{1}{\pi\sigma_z^2}\sqrt{2\sigma_z^2-z^2}1πJ1(2κ)2κ\frac{1}{\pi}\frac{J_1(\sqrt{2}\kappa)}{\sqrt{2}\kappa}

Table 2. D(κ)D(\kappa) for several longitudinal distributions. h(x)h(x) is the step function. The beam size is defined by σz=εzβz\sigma_z=\sqrt{\varepsilon_z\beta_z}, and particles are distributed over 0≤J^≤10\leq\hat{J}\leq1 and ∣z∣≤2σz|z|\leq\sqrt{2}\sigma_z in the water-bag distribution, where J^=J/εz=r2/2\hat{J}=J/\varepsilon_z=r^2/2.

Here, variables are normalized as κ=ωσz/(cβ)\kappa=\omega\sigma_z/(c\beta), r=2J/εzr=\sqrt{2J/\varepsilon_z}. DD in the first line of Eq. (47) is defined as:

D(κ)=∫0∞r′ dr′J0(κr′)ψ(r′).(48)D(\kappa)=\int_0^\infty r'\,dr'J_0(\kappa r')\psi(r'). \tag*{(48)}

The integrations have been carried out for several longitudinal distributions, and the results for the Gaussian and water-bag distributions are summarized in Ref. [10].

In the second line of Eq. (47), κ−χ\kappa-\chi enters the integrand, but by shifting the integration over ω\omega, it can be written in a form independent of χ\chi. This is possible because the impedance of the space-charge force is constant.

Here, as in Eq. (34), we express it as an equation of motion for vlv_l:

dvl(J)ds+2πCi(νx+lνz)vl(J)=Zscβx2C∑l′il−l′+1[vl′(J)∫dωJl−l′(κr)D(κ)−∫dJ′ψ(J′)vl′(J′)∫dωJl(κr)Jl′(κr′)],(49)\frac{dv_l(J)}{ds}+\frac{2\pi}{C}i(\nu_x+l\nu_z)v_l(J) =\frac{Z_{sc}\beta_x}{2C}\sum_{l'}i^{l-l'+1}\left[v_{l'}(J)\int d\omega J_{l-l'}(\kappa r)D(\kappa)-\int dJ'\psi(J')v_{l'}(J')\int d\omega J_l(\kappa r)J_{l'}(\kappa r')\right], \tag*{(49)}

Because the change due to the space-charge force is equally distributed between xlx_l and plp_l, a factor of 1/21/2 appears. Setting vl(J)∝e−2πiνs/Cv_l(J)\propto e^{-2\pi i\nu s/C}:

(ν−νx−lνz)vl(J)=−Zscβx4π∑l′il−l′[vl′(J)∫dωJl−l′(κr)D(κ)−∫dω∫dJ′ψ(J′)vl′(J′)Jl(κr)Jl′(κr′)],(50)(\nu-\nu_x-l\nu_z)v_l(J) =-\frac{Z_{sc}\beta_x}{4\pi}\sum_{l'}i^{l-l'}\left[v_{l'}(J)\int d\omega J_{l-l'}(\kappa r)D(\kappa)-\int d\omega\int dJ'\psi(J')v_{l'}(J')J_l(\kappa r)J_{l'}(\kappa r')\right], \tag*{(50)}

By solving this equation as an eigenvalue problem, the eigen-oscillation modes and their eigenfrequencies can be found.

Eq. (50) is reduced to an eigenvalue problem by discretizing in JJ. Actually, we discretize with r=2J/εzr=\sqrt{2J/\varepsilon_z} (rk=kΔrr_k=k\Delta r) instead of JJ. The orthogonal functions hk(r)h_k(r) are expressed as

hk(r)=1/Δrfor∣r−rk∣≤Δr/2(51)h_k(r)=1/\Delta_r\quad\text{for}\quad|r-r_k|\leq\Delta_r/2 \tag*{(51)}

and hk(r)=0h_k(r)=0 elsewhere. The normalization and orthogonality relations are:

∫dr hk(r)=1,∫0∞dr hi(r)hk(r)=δikΔr.(52)\int dr\,h_k(r)=1,\qquad\int_0^\infty dr\,h_i(r)h_k(r)=\frac{\delta_{ik}}{\Delta_r}. \tag*{(52)}

This method was first used by K. Oide for the analysis of longitudinal instabilities [11] and extended to the transverse beam-beam instabilities [12].

Before solving the eigenvalue problem, we try to calculate the tune shift as the lowest-order perturbation, taking the diagonal elements of Eq. (50). As explained in detail later, taking the discretization as the basis of the eigenvalue problem, the first term of the diagonal element on the right side of Eq. (50) is larger than the second term. The JJ-dependent tune shift is:

Δνx,l(r)=−kscβxC4πσz∫−∞∞dκ J0(κr)D(κ)(53)\Delta\nu_{x,l}(r)=-\frac{k_{sc}\beta_x C}{4\pi\sigma_z}\int_{-\infty}^{\infty}d\kappa\,J_0(\kappa r)D(\kappa) \tag*{(53)}

The tune shift, which is independent of ll, is given for the Gaussian distribution as

Δνl(r)=Δνsce−r2/4I0(r24),Δνsc=−βx4πkscC2πσz.(54)\Delta\nu_l(r)=\Delta\nu_{sc}e^{-r^2/4}I_0\left(\frac{r^2}{4}\right),\qquad\Delta\nu_{sc}=-\frac{\beta_x}{4\pi}\frac{k_{sc}C}{\sqrt{2\pi}\sigma_z}. \tag*{(54)}

That for the water-bag distribution is

Δνl(r)=Δνsc∫0∞J0(κr)J1(2κ)dκκ,\Delta\nu_l(r)=\Delta\nu_{sc}\int_0^\infty J_0(\kappa r)J_1(\sqrt{2}\kappa)\frac{d\kappa}{\kappa},
Δνsc=−βx4π2kscCπσz.(55)\Delta\nu_{sc}=-\frac{\beta_x}{4\pi}\frac{\sqrt{2}k_{sc}C}{\pi\sigma_z}. \tag*{(55)}

The coefficient Δνsc\Delta\nu_{sc} is the so-called space-charge tune shift at the peak line density ρ(0)\rho(0). Figure 5 shows the rr-dependent term, Δνl(r)/Δνsc\Delta\nu_l(r)/\Delta\nu_{sc}, given in Eqs. (54) and (55).

Figure 5. Plot of Δνl(r)/Δνsc\Delta\nu_l(r)/\Delta\nu_{sc} for the Gaussian and water-bag distributions.

Next, let us move on to solving the eigenvalue problem. Expand vx(J,ϕ)v_x(J,\phi) in terms of hk(r)h_k(r) and represent the coupling coefficient as vlkv_{lk}:

vl(J)=∑kvlkhk(r)(56)v_l(J)=\sum_k v_{lk}h_k(r) \tag*{(56)}

Applying Δr∫dr hk(r)\Delta_r\int dr\,h_k(r) to Eq. (50) yields:

(ν−νx−lνz)vlk=−Zscβx4π∑l′il−l′[vl′k′∫dω Jl−l′(κrk)D(κ)−∑k′rk′ψ(rk′)Δrvl′k′′∫dω Jl(κrk)Jl′(κrk′)].(57)\begin{aligned} (\nu-\nu_x-l\nu_z)v_{lk}={}&-\frac{Z_{\mathrm{sc}}\beta_x}{4\pi}\sum_{l'}i^{l-l'}\left[v'_{l'k}\int d\omega\,J_{l-l'}(\kappa r_k)D(\kappa)\right.\\ &\left.-\sum_{k'}r_{k'}\psi(r_{k'})\Delta_rv'_{l'k'}\int d\omega\,J_l(\kappa r_k)J_{l'}(\kappa r_{k'})\right]. \tag*{(57)} \end{aligned}

With the change of variables from JJ to rr, the distribution function will henceforth be written as ψ(r)\psi(r), with the two related by ψ(J)dJ=ψ(r)r dr\psi(J)dJ=\psi(r)r\,dr.

The eigenvalue equation to be solved is:

(νx+lνz+M)v=νv.(58)(\nu_x+l\nu_z+M)v=\nu v. \tag*{(58)}

The first two terms on the left side are the diagonal matrix (νx+lνz)δll′δkk′(\nu_x+l\nu_z)\delta_{ll'}\delta_{kk'}, and the third term, the matrix MM, is expressed as:

Mlkl′k′=−Zscβx4πil−l′[δkk′∫dω Jl−l′(κrk)D(κ)−rk′ψ(rk′)Δr∫dω Jl(κrk)Jl′(κrk′)].(59)M_{lkl'k'}=-\frac{Z_{\mathrm{sc}}\beta_x}{4\pi}i^{l-l'}\left[\delta_{kk'}\int d\omega\,J_{l-l'}(\kappa r_k)D(\kappa)-r_{k'}\psi(r_{k'})\Delta_r\int d\omega\,J_l(\kappa r_k)J_{l'}(\kappa r_{k'})\right]. \tag*{(59)}

Equations (58) and (59) constitute a complex square matrix of dimension (2lmax⁡+1)nJ(2l_{\max}+1)n_J. Here, we introduce a cutoff for the frequency in the integral. Since rr was discretized with Δr\Delta_r, we used κc=1/Δr\kappa_c=1/\Delta_r, i.e. ωc=2cβ/(nJσz)\omega_c=\sqrt{2c\beta}/(n_J\sigma_z), as the cutoff frequency and set it (±ωc)(\pm\omega_c) as the upper and lower limits of the ω\omega integral.

Since this matrix is not Hermitian, the eigenvectors satisfy the following orthogonality relation, represented by left and right eigenvectors:

∑lkvlk(m,L)∗vlk(m′,R)=δmm′.(60)\sum_{lk}v_{lk}^{(m,L)*}v_{lk}^{(m',R)}=\delta_{mm'}. \tag*{(60)}

The eigen-oscillation states are represented by a linear combination of the basis vectors consisting of the radial function hk(r)h_k(r) and the azimuthal function eilϕ+iχz/σze^{il\phi+i\chi z/\sigma_z}:

v(J,ϕ)m=∑lkvlk(m,R)hk(r)eilϕ+iχz/σz(61)v(J,\phi)_m=\sum_{lk}v_{lk}^{(m,R)}h_k(r)e^{il\phi+i\chi z/\sigma_z} \tag*{(61)}
v(J,ϕ)m†=∑lkvlk(m,L)∗hk(r)e−ilϕ−iχz/σz.(62)v(J,\phi)_m^\dagger=\sum_{lk}v_{lk}^{(m,L)*}h_k(r)e^{-il\phi-i\chi z/\sigma_z}. \tag*{(62)}

The state corresponding to the conjugate is defined as v(J,ϕ)m†v(J,\phi)_m^\dagger. The orthogonality relation for the eigen-oscillation states is:

12π∫dr dϕ v(J,ϕ)m′†v(J,ϕ)m=δmm′Δr.(63)\frac{1}{2\pi}\int dr\,d\phi\,v(J,\phi)_{m'}^\dagger v(J,\phi)_m=\frac{\delta_{mm'}}{\Delta_r}. \tag*{(63)}

Here, orthogonality in ll and the orthogonality relations for k,mk,m, Eqs. (52) and (60), were used. The time evolution of the oscillation state involves a phase rotation by the eigenfrequency (tune) at each turn:

v(J,ϕ;t)m=e−2πiνmtv(J,ϕ;0)m.(64)v(J,\phi;t)_m=e^{-2\pi i\nu_m t}v(J,\phi;0)_m. \tag*{(64)}

Any oscillation state can be represented by a linear combination of eigenvectors:

v(J,ϕ)=∑mamv(J,ϕ)m(65)v(J,\phi)=\sum_m a_m v(J,\phi)_m \tag*{(65)}

Given an oscillation state v(J,ϕ)v(J,\phi), the coefficients of the linear combination ama_m can be determined by applying (2π)−1∫dr dϕ v(J,ϕ)m†(2\pi)^{-1}\int dr\,d\phi\,v(J,\phi)_m^\dagger to both sides:

am=Δr2π∫dr dϕ v(J,ϕ)m†v(J,ϕ)=Δr2π∫dϕ∑lkvlk(m,L)∗e−ilϕ−iχzk/σzv(Jk,ϕ).(66)\begin{aligned} a_m=\frac{\Delta_r}{2\pi}\int dr\,d\phi\,v(J,\phi)_m^\dagger v(J,\phi) \\ &=\frac{\Delta_r}{2\pi}\int d\phi\sum_{lk}v_{lk}^{(m,L)*}e^{-il\phi-i\chi z_k/\sigma_z}v(J_k,\phi). \tag*{(66)} \end{aligned}

Beam oscillation mode given by solving the eigenvalue problem

The eigenvalue problem was solved using the parameters of the CSNS, where the longitudinal phase-space distribution was Type W in Table II. This is a uniform-density distribution in phase space, a so-called water-bag distribution. Using a vector of length 40×41=164040\times41=1640, with nJ=40n_J=40, lmax⁡=20l_{\max}=20, we solve the eigenvalue problem for a 1640×16401640\times1640 matrix for each bunch intensity.

Figure 6 plots the eigen-tunes as a function of the bunch population, where the number of particles in the bunch was varied up to Np=1×1013N_p=1\times10^{13} in 100 steps, and the eigenvalue problem was solved at each step. Without space-charge forces, the eigen-tunes were νx+lνz\nu_x+l\nu_z and were degenerate for the radial modes vl(J)v_l(J), but the space-charge force lifts this degeneracy, generating a tune shift for each radial mode. Particles with smaller synchrotron amplitudes experience a larger negative tune shift due to stronger space-charge forces, and vice versa. Many of the lines shifting negatively with increasing NpN_p in the figure are due to the space-charge tune shift. The magnitude of the shift corresponds to the synchrotron amplitude (rr or JJ).

Tunes of oscillations for each bunch intensity

Figure 6. Tunes of oscillations for each bunch intensity, obtained by solving the eigenvalue problem, where lmax⁡=10l_{\max}=10 and nJ=40n_J=40. The horizontal scale of the bunch population is Np=0N_p=0–1×10131 \times10^{13} in (a) and is enlarged to 00–1×10121 \times10^{12} in (b).

In Figure 6, for l≥0l\geq0, there are modes clearly isolated from other modes. For better readability of the figure, lmax⁡l_{\max} is set to 10. This isolated mode is easily understood. When the chromaticity is 0, a distribution where all particles in the bunch are uniformly shifted undergoes betatron oscillations without being affected by space-charge forces. For l<0l<0, it is not clear whether there are isolated modes.

Figure 7 shows the eigen-oscillation patterns of the separated modes represented by Eq. (61). The top row is for Np=0.44×1012N_p=0.44\times10^{12}, arranged from left to right for l=1,2,3l=1,2,3. Only the real part is plotted, so the amplitude distribution x(J,ϕ;0)x(J,\phi;0) is depicted. The distributions show vertical oscillations corresponding to l=1,2,3l=1,2,3 in the azimuthal direction. Tracking the time evolution by multiplying by e−iωte^{-i\omega t} reveals the oscillation pattern rotating in the zz-δ\delta phase space due to synchrotron oscillations.

Eigen-oscillation patterns of the separated modes

Figure 7. Eigen-oscillation patterns of the separated modes obtained from Eq. (61) using the eigenvectors. Horizontal and vertical axes are zz and δ\delta, respectively. Top row: Np=0.44×1012N_p=0.44 \times10^{12}, ∣Δνsc/νz∣=1.6\lvert\Delta\nu_{sc}/\nu_z\rvert=1.6, Bottom row: Np=3.5×1012N_p=3.5 \times10^{12}, ∣Δνsc/νz∣=12\lvert\Delta\nu_{sc}/\nu_z\rvert=12. From left to right: l=1,2,3l=1, 2, 3.

The bottom row shows the oscillation patterns for Np=3.5×1012N_p=3.5\times10^{12}. They are uniform in the δ\delta direction, and the structure is only seen in the zz direction. Space-charge forces do not act on such a distribution as seen in Eq. (40). Individual particles undergo synchrotron oscillations. Space-charge forces act so that synchrotron oscillations are not coherently visible in the direction of synchrotron motion. When tracking the time evolution, synchrotron oscillations are not seen when the amplitude is large, but rotation becomes visible when the amplitude becomes small. The tune approaches νx\nu_x but does not completely reach νx\nu_x, so the synchrotron oscillation

component has not vanished.

Decoherence and recoherence of a kicked beam

Consider a distribution uniformly shifted in the xx direction. The initial state is given by v(J,ϕ;0)=x0v(J,\phi;0)=x_0. This initial distribution is expanded in terms of the eigenvectors:

x0=∑mamv(J,ϕ)m.(67)x_0=\sum_m a_m v(J,\phi)_m. \tag*{(67)}

Substituting into Eq. (66), the coefficients ama_m are obtained for x0=1x_0=1 as:

am=Δr∑lkvlk(m,L)∗i−lJl(χrk).(68)a_m=\Delta_r \sum_{lk} v_{lk}^{(m,L)*} i^{-l} J_l(\chi r_k). \tag*{(68)}

The extent to which each eigenmode is included in the uniformly shifted distribution can be seen from ama_m. Figure 8 shows the amplitudes of the eigenmodes for each chromaticity χ\chi. At χ=0\chi=0, only the l=0l=0 mode with tune equal to νx\nu_x is induced. For χ>0\chi>0, as χ\chi increases, modes with tunes higher than νx\nu_x become included. Modes with tunes lower than νx\nu_x are continuously distributed as seen in Figure 6, so individual modes have small components.

Linear combination coefficients

Figure 8. Linear combination coefficients ama_m for the beam initialized by a uniform dipole distribution, obtained from Eq. (68).

Each eigenmode oscillates according to its own eigen-tune νm\nu_m:

v(J,ϕ;t)=∑mamv(J,ϕ)me−2πiνmt.(69)v(J,\phi;t)=\sum_m a_m v(J,\phi)_m e^{-2\pi i\nu_m t}. \tag*{(69)}

Figure 9 plots the centroid oscillation of the bunch for each chromaticity. Coherence is lost due to differences in the tunes of the included eigenmodes, and the centroid initially approaches zero, but coherence recovers again. The recovery period is 1/νz=1001/\nu_z=100 turns, as seen in Figure 2 without the space-charge force, but with the space-charge force, the period becomes longer than 1/νz1/\nu_z. Although the l=1l=1 mode is quite close to the tune, the modes included vary with chromaticity as seen in Figure 8, so the degree of deviation of the period from 1/νz1/\nu_z depends on chromaticity. For small chromaticity, a small number of modes νx<ν<νx+νz\nu_x < \nu< \nu_x + \nu_z are excited, so the periodicity is clearly longer than 1/νz1/\nu_z. For high chromaticity, several to tens of modes with ν>νx\nu> \nu_x are excited, so the behavior is complex, though the periodicity is longer than 1/νz1/\nu_z.

Figure 9. Centroid oscillation of the bunch for each chromaticity, with an initial amplitude of x0=1x_0 = 1, obtained from Eq. (69), where Np=3.5×1012N_p = 3.5 \times10^{12}, ∣Δνsc/νz∣=12\lvert\Delta\nu_{sc}/\nu_z\rvert= 12.

Frequency response of forced oscillation: Beam Transfer Function

For the case with a strong space-charge force, we find the response of each eigenmode to a sinusoidal forced oscillation Δpx(J,ϕ)=Aeiωz/(cβ)\Delta p_x(J,\phi) = A e^{i\omega z/(c\beta)}. The momentum change at s=0s = 0 is given by

Δνx(J,ϕ)=∑mΔamv(J,ϕ)m=iAeiωz/(cβ).(70)\Delta\nu_x(J,\phi) = \sum_m \Delta a_m v(J,\phi)_m = i A e^{i\omega z/(c\beta)}. \tag*{(70)}

Multiplying by v(J,ϕ)m†v(J,\phi)_m^\dagger and integrating over J,ϕJ,\phi, the coefficient Δam\Delta a_m for each eigenmode is obtained:

Δam=i2π∫dr∫dϕ Av(J,ϕ)m†eiωz/(cβ)=iAΔr2π∫dϕ∑lkvlk(m,L)∗e−ilϕ−iχzk/σz+iωzk/(cβ)=AΔr∑lkvlk(m,L)∗il+1Jl((κ−χ)rk).(71)\begin{aligned} \Delta a_m &= \frac{i}{2\pi}\int dr\int d\phi\,A v(J,\phi)_m^\dagger e^{i\omega z/(c\beta)}\\ &= \frac{i A\Delta r}{2\pi}\int d\phi\sum_{lk}v_{lk}^{(m,L)*}e^{-il\phi-i\chi z_k/\sigma_z+i\omega z_k/(c\beta)}\\ &= A\Delta r\sum_{lk}v_{lk}^{(m,L)*}i^{l+1}J_l((\kappa-\chi)r_k). \tag*{(71)} \end{aligned}

Considering the repetition over ss, we express the change in ama_m by a differential equation:

damds+2πiCνmam=AΔr∑lkil+1vlk(m,L)∗Jl((κ−χ)rk)∑ne−2πinνδ(s−nC)(72)\frac{d a_m}{d s}+\frac{2\pi i}{C}\nu_m a_m =A\Delta r\sum_{lk}i^{l+1}v_{lk}^{(m,L)*}J_l((\kappa-\chi)r_k)\sum_n e^{-2\pi i n\nu\delta(s-nC)} \tag*{(72)}

The solution of Eq. (72) is obtained similarly to Eqs. (35)–(36), giving:

am(ν)=−AΔr∑p∑lkilvlk(m,L)∗Jl((κ−χ)rk)ν′−νme−2πiν′t.(73)a_m(\nu)=-A\Delta r\sum_p\frac{\sum_{lk}i^l v_{lk}^{(m,L)*}J_l((\kappa-\chi)r_k)}{\nu'-\nu_m}e^{-2\pi i\nu't}. \tag*{(73)}

When the beam receives the momentum change at a certain frequency, the strength of each induced mode is indicated by ama_m. The condition for the linear combination coefficient ama_m corresponding to the modal response to form a peak is that the denominator of Eq. (73) becomes zero: ν′−νm=0\nu'-\nu_m=0. The condition means that excitation occurs when the oscillation mode tune νm\nu_m and the forced oscillation tune ν\nu differ by an integer pp. As is done in Sec. II C, we express νm\nu_m using the fractional part. Consider the case where a certain mode mm is excited by a sweep of the input frequency (ν\nu). The tune (frequency) at which am(ν)a_m(\nu) peaks corresponds to multiple integer values pp, with frequencies νm,p=νm+p\nu_{m,p}=\nu_m+p and κm,p=2πνm,pσz/C\kappa_{m,p}=2\pi\nu_{m,p}\sigma_z/C. Denote the response at a specific pp in the sum over pp in Eq. (73) as am,pa_{m,p}:

am,p≡∑lkilvlk(m,L)∗Jl((κm,p−χ)rk).(74)a_{m,p}\equiv\sum_{lk}i^l v_{lk}^{(m,L)*}J_l((\kappa_{m,p}-\chi)r_k). \tag*{(74)}

Figure 10 presents the amplitude am,pa_{m,p} induced by forced oscillation for χ=0\chi=0, where the horizontal axis is the fractional part of the oscillation tune and different pp are overlaid. Points are located at νm\nu_m on the horizontal axis. The eigenvectors are independent of χ\chi and pp. As seen in Eq. (74), ama_m depends on χ\chi and pp through the argument of the Bessel function, (κm,p−χ)rk(\kappa_{m,p}-\chi)r_k. The figure does not change for a finite chromaticity except for a shift in pp, Δp=Cχ/(2πσz)=ξ/ηp\Delta p=C\chi/(2\pi\sigma_z)=\xi/\eta_p.

Figure 10. Amplitude ∣am∣|a_m| induced by forced oscillation, where the beam intensity is Np=(7.5,2.5,1.0,N_p=(7.5,2.5,1.0, and 0.278)×10120.278)\times10^{12} (Δνsc=0.27,0.09,0.036,0.01,\Delta\nu_{\mathrm{sc}}=0.27,0.09,0.036,0.01, and νz=0.01\nu_z=0.01) in plots (a)–(d).

The mode induced at the frequency corresponding to νm,p\nu_{m,p}, combined with the eigenvector, has the following dipole amplitude distribution:

am,pv(J,ϕ)m=am,p∑lkvlk(m,R)hk(r)eilϕ+iχz/σz.(75)a_{m,p}v(J,\phi)_m=a_{m,p}\sum_{lk}v_{lk}^{(m,R)}h_k(r)e^{il\phi+i\chi z/\sigma_z}. \tag*{(75)}

By integrating Eq. (75) over δ\delta or (z,δ)=(J,ϕ)(z,\delta)=(J,\phi) for each pp and mm, measurable quantities, ρx(z)\rho_x(z) and ⟨v⟩\langle v\rangle are obtained. They characterize the magnitude of the response of each mode to the input frequency. The response for ⟨v⟩\langle v\rangle, in which mode mm excited by frequency ν=νm+p\nu=\nu_m+p, is:

⟨vm,p⟩=am,p∫r dr dϕ ψ(r)v(J,ϕ)m=2πam,p∑lkilvlk(m,R)Jl(χrk)rkψ(rk).\begin{aligned} \langle v_{m,p}\rangle&= a_{m,p}\int r\,\mathrm{d}r\,\mathrm{d}\phi\,\psi(r)v(J,\phi)_m \\ &=2\pi a_{m,p}\sum_{lk}i^l v_{lk}^{(m,R)}J_l(\chi r_k)r_k\psi(r_k). \end{aligned}

Figure 11 shows the response for ⟨vm,p⟩\langle v_{m,p}\rangle at Np=7.5×1012N_p=7.5\times10^{12}. At χ=0\chi=0, the result is almost the same as without the space-charge force. Only the betatron mode ν′=0.85=ν\nu'=0.85=\nu responds strongly for several pp near p=0p=0. No dipole moment is associated with the oscillation of the l≠0l\ne0 modes even in the presence of the space-charge force. The difference is clear for χ>0\chi>0 when compared to Figure 3 without the space-charge force. A peak appears at l>0l>0 for χ>0\chi>0, and the peak shifts to higher ll for increasing χ\chi. This behavior is the same for χ<0\chi<0: the peak structure depends only on ∣χ∣|\chi|. The height of the response is higher than that without the space-charge force. There is also no response for the l<0l<0 modes, as seen in Figure 8 and Figure 10.

Relation between output tune and amplitude for each chromaticity

Figure 3. Relation between output tune and amplitude for each chromaticity (without space-charge forces), where νx′=0.85\nu'_x=0.85 and νz=0.01\nu_z=0.01.

The dipole moment as a function of zz is given by the integral of Eq. (75) over δ\delta,

ρx(z)m,p=am,p∫dδ v(J,ϕ)mψ(J).(76)\rho_x(z)_{m,p}=a_{m,p}\int\mathrm{d}\delta\,v(J,\phi)_m\psi(J). \tag*{(76)}

v(J,ϕ)mv(J,\phi)_m and its δ\delta integral, ρx(z)m\rho_x(z)_m, exhibit a sinusoidal structure in zz due to the chromatic modulation eiχz/σze^{i\chi z/\sigma_z} and the oscillation mode subject to the space-charge force, as shown in Figure 7. The Fourier component of the dipole moment at the input frequency ω=ωm,p\omega=\omega_{m,p} is given by

ρx(ωm,p)=∫dz ρx(z)m,pe−iωm,pz/(cβ)=am,p∫r dr dϕ∑lkvlk(m,R)hk(r)ψ(J)eilϕ+iχz/σz−iωm,pz/(cβ)=2πam,p∑lki−lvlk(m,R)Jl((κm,p−χ)rk)rkψ(rk).(77)\begin{aligned} \rho_x(\omega_{m,p}) &= \int\mathrm{d}z\,\rho_x(z)_{m,p}e^{-i\omega_{m,p}z/(c\beta)} \\ &=a_{m,p}\int r\,\mathrm{d}r\,\mathrm{d}\phi\sum_{lk}v_{lk}^{(m,R)}h_k(r)\psi(J)e^{il\phi+i\chi z/\sigma_z-i\omega_{m,p}z/(c\beta)} \\ &=2\pi a_{m,p}\sum_{lk}i^{-l}v_{lk}^{(m,R)}J_l((\kappa_{m,p}-\chi)r_k)r_k\psi(r_k). \tag*{(77)} \end{aligned}

Figure 11. Beam response for ⟨vm,p⟩\langle v_{m,p}\rangle induced by a forced oscillation with frequency νm,p\nu_{m,p}.

Figure 12 shows the response of ρx(ωm,p)\rho_x(\omega_{m,p}) at Np=7.5×1012N_p=7.5\times10^{12}. The corresponding figure without the space-charge force is shown in Figure 4. Plot (a) depicts the response at the eigen-tune νm\nu_m, and those for different pp are overlaid. The plot is drawn for χ=0\chi=0; the same plot applies for ∣χ∣≠0|\chi|\ne0. The discrete mode with visible amplitude corresponds to the isolated mode seen in Figure 6. The mode ν=0.85\nu=0.85 is the isolated pure dipole mode. The modes ν>0.85\nu>0.85 are lined up from left to right as l=1,2,3,…l=1,2,3,\ldots. We express the isolated modes using ll. ρx(ωm,p)\rho_x(\omega_{m,p}) for the l≥0l\ge0 isolated modes is shown in plot (b), where χ=0\chi=0. As seen in Figure 4, plot (c) shifts as χ\chi changes, by Δp=CΔχ/(2πσz)=Δξ/ηp\Delta p=C\Delta\chi/(2\pi\sigma_z)=\Delta\xi/\eta_p. This is because Eq. (39) contains Jl(κm,p−χ)2J_l(\kappa_{m,p}-\chi)^2. Here, both Eqs. (74) and (77) contain Jl(κm,p−χ)J_l(\kappa_{m,p}-\chi). The eigenvectors vlkv_{lk} do not depend on χ\chi, because the space-charge impedance in Eq. (44) is independent of frequency. The peak structure of ρx(ωm,p)\rho_x(\omega_{m,p}) only shifts with pp, but its values do not change. This explains why the same plot is obtained for ∣χ∣≠0|\chi|\ne0 in plot (a). Plot (c) shows the shift of the response ρx(ωm,p)\rho_x(\omega_{m,p}) for different chromaticities.

Beam response plots

Figure 4. Beam response ρx(ωl,p)\rho_x(\omega_{l,p}) for forced oscillation at the frequency ωl,p\omega_{l,p} (without space-charge forces). (a) the response as a function of the fractional input tune ν′\nu' for various χ\chi, with different pp overlaid, (b) the response as a function of pp for each ll at ξ=0\xi=0, and (c) the response for different χ\chi in the l=1l=1 mode.

The difference from the response without the space-charge force is (i) there is a response for only one mode for the l≥0l\ge0 modes, but no response for the l<0l<0 modes, (ii) the frequency separation of the response between ll modes is narrower than νz\nu_z, and (iii) the response for the l>0l > 0 modes is more than twice as strong as that without the space-charge force.

This behavior is closely connected to the oscillation mode with and without the space-charge force. (i) and (ii) are explained by the fact that an isolated mode exists for each of the l≥0l \ge0 modes, whereas no isolated modes exist for l<0l < 0. The modes with l<0l < 0 have only a continuous spectrum, as shown in Figure 6. (iii) is explained by the mode pattern seen in Figure 7. The distribution of the transverse amplitude ∑lκvlk(m)Jl(κrk)eilϕ\sum_l \kappa v_{lk}^{(m)} J_l(\kappa r_k)e^{il\phi} is sinusoidal in zz, and is independent of δ\delta. The Fourier integral therefore gives a larger value compared with the pure synchro-betatron mode, eilϕe^{il\phi}.

The continuous modes do not respond to forced oscillations: this is due to Landau damping. Even for the l>0l > 0 modes, the response is limited to the discrete modes. Only isolated modes separated from the continuum respond to forced oscillation. Figure 13 summarizes the frequency response ρx(ωm,p)\rho_x(\omega_{m,p}) for the isolated ll modes for various bunch intensities, where ∣Δνsc∣=νz=0.01\lvert\Delta\nu_{sc}\rvert= \nu_z = 0.01 at Np=0.28×1012N_p = 0.28 \times10^{12}. The lines with ∣Δνsc/νz∣≥9\lvert\Delta\nu_{sc}/\nu_z\rvert\ge9 are almost overlapping, and those with ∣Δνsc/νz∣≤1\lvert\Delta\nu_{sc}/\nu_z\rvert\le1 are almost overlapping. ∣Δνsc/νz∣∼3\lvert\Delta\nu_{sc}/\nu_z\rvert\sim3 shows transient behavior. For ∣Δνsc/νz∣≥9\lvert\Delta\nu_{sc}/\nu_z\rvert\ge9 and ≤1\le1, the difference in the response gets bigger as ll increases, and at l=3l = 3 and 4, it differs by a factor of five.

Figure 13. Frequency response Max(ρx(ωm,p))Max(\rho_x(\omega_{m,p})) for the isolated ll modes for various bunch intensities, Np=(7.5,5.0,2.5,1.0,0.28,0.1)×1012N_p = (7.5, 5.0, 2.5, 1.0, 0.28, 0.1) \times10^{12}, where −Δνsc=νz=0.01-\Delta\nu_{sc} = \nu_z = 0.01 at Np=0.28×1012N_p = 0.28 \times10^{12}.

Beam Instability

We discuss the beam instability that arises when the frequency components possessed by the beam match those of the impedance. Frequency shifts caused by the impedance are assumed to be small: that is, transverse mode coupling is not considered. The main subjects are the head-tail instability and the coupled-bunch instability. The impedance is expressed as a resonator model with Q≳10Q \gtrsim10, and the contribution of the kick factor is small relative to the synchrotron tune. Cases where the resonator frequency is high with a large RsR_s but not a large Rs/QR_s/Q are also included. This would apply to the impedance and beam instability in high-intensity proton accelerators such as neutron sources. The impedance induced by the ceramic chambers in CSNS is Zpeak=1 MΩ/mZ_{\mathrm{peak}} = 1\ \mathrm{M}\Omega/\mathrm{m} at fR=123 kHzf_R = 123\ \mathrm{kHz} with Q=40Q = 40 [13]. The impedance induced by the kicker magnets in J-PARC is Zpeak∼160 kΩ/mZ_{\mathrm{peak}} \sim160\ \mathrm{k}\Omega/\mathrm{m} at fR=0.4f_R = 0.4–4 MHz4\ \mathrm{MHz} with Q=30Q = 30–5050 [14].

In conventional instability theories, the underlying oscillation modes were those of the free synchro-betatron oscillation in Eq. (22). The synchro-betatron modes are degenerate in the radial direction, so there is uncertainty about which radial mode to use as the base. Generally, the lowest-order flat radial mode is used. It has been found that the oscillation modes differ significantly for beams with strong space-charge forces. The eigenvalue

problem including space-charge forces and the impedance was solved in Ref. [10], where the impedance was evaluated at νx\nu_x. This implicitly assumed Q<νR/νzQ < \nu_R/\nu_z, where νR=fR/f0\nu_R = f_R/f_0 was the resonator frequency normalized by the revolution frequency. Here, following the conditions described earlier, we take the underlying eigenmodes as oscillation modes that account for the space charge. The space-charge force dominates the effects on intra-bunch oscillations, and the instability is caused by the impedance under resonance conditions in which the oscillation mode frequency and the impedance frequency match.

The wake force Wx(z)W_x(z) is now added to the equation of motion. In this section, we use the wake and impedance in conventional units Wx[m−2]W_x[\mathrm{m}^{-2}] and Zx[s/m2]Z_x[\mathrm{s}/\mathrm{m}^2], respectively [15]. The equation is expressed in terms of its Fourier transform, the impedance Zx(ω)Z_x(\omega).

dv(J,ϕ;s)mds+2πiνmCv(J,ϕ;s)m=−iNprpβx2γC∑n=−∞∞∫Wx(z−z′−nC)v(J′,ϕ′;s)me2πinνmψ(J′)dJ′dϕ′\frac{d v(J,\phi;s)_m}{d s}+\frac{2\pi i\nu_m}{C}v(J,\phi;s)_m=-\frac{iN_pr_p\beta_x}{2\gamma C}\sum_{n=-\infty}^{\infty}\int W_x(z-z'-nC)v(J',\phi';s)_m e^{2\pi i n\nu_m\psi(J')}dJ'd\phi'

Using v(J,ϕ;s)m=e−2πiνs/Cv(J,ϕ;0)mv(J,\phi;s)_m=e^{-2\pi i\nu s/C}v(J,\phi;0)_m, multiplying both sides by v(J,ϕ)m†v(J,\phi)_m^\dagger and integrating over (r,ϕ)(r,\phi):

(ν−νm)=−iNprpβx4πγω0∑p=−∞∞Z(ωmp)gm,L∗(κmp−χ)gm,R(κmp−χ)(78)(\nu-\nu_m)=-\frac{iN_pr_p\beta_x}{4\pi\gamma}\omega_0\sum_{p=-\infty}^{\infty}Z(\omega_{mp})g_{m,L}^{*}(\kappa_{mp}-\chi)g_{m,R}(\kappa_{mp}-\chi) \tag*{(78)}

where

gm,L∗(κ)=∑lkilvlk(m,L)∗Jl(κrk),g_{m,L}^{*}(\kappa)=\sum_{lk}i^l v_{lk}^{(m,L)*}J_l(\kappa r_k),
gm,R(κ)=∑lki−lvlk(m,R)Jl(κrk)rkψ(rk)Δr.g_{m,R}(\kappa)=\sum_{lk}i^{-l}v_{lk}^{(m,R)}J_l(\kappa r_k)r_k\psi(r_k)\Delta r.

The growth rate of mode mm per revolution can be determined by examining the imaginary part of the tune multiplied by 2π2\pi:

Gr=−Nprpβx2ω0Re⁡[∑pZx(ωm,p)gm,L∗(κm,p−χ)gm,R(κm,p−χ)](79)G_r=-\frac{N_pr_p\beta_x}{2}\omega_0\operatorname{Re}\left[\sum_p Z_x(\omega_{m,p})g_{m,L}^{*}(\kappa_{m,p}-\chi)g_{m,R}(\kappa_{m,p}-\chi)\right] \tag*{(79)}

gm,L∗(κm,p−χ)gm,R(κm,p−χ)g_{m,L}^{*}(\kappa_{m,p}-\chi)g_{m,R}(\kappa_{m,p}-\chi) in Eq. (79) contains the same terms as ρx(ωm,p)\rho_x(\omega_{m,p}) expressed by Eqs. (74) and (77). Thus, a rough sketch of the instability can be understood from Figure 4, Figure 12 and Figure 13. For example, for Re⁡[Z(ω)]=1 MΩ/m\operatorname{Re}[Z(\omega)]=1\ \mathrm{M}\Omega/\mathrm{m}, the growth rate and time are 5×10−45\times10^{-4} and 4 ms4\ \mathrm{ms}, when νR=−νx+p\nu_R=-\nu_x+p. The growth rate is positive when the impedance is negative: that is, ω<0\omega<0. The condition and the result do not change with and without the space-charge force because of l=0l=0, χ=0\chi=0. For a finite chromaticity, differences emerge between the cases with and without the space-charge force. Without the space-charge force, instability occurs at −νR=νx+lνz+p−ξ/ηp-\nu_R=\nu_x+l\nu_z+p-\xi/\eta_p. The growth rate decreases to 27%, 16%, 12% and 10% for ∣l∣=1,2,3\lvert l\rvert=1,2,3 and 4 as shown in Figure 4(b). The growth rate is independent of the sign of ll.

Response for forced oscillation of the beam subject to the space-charge force

Figure 12. Response for forced oscillation of the beam subject to the space-charge force. (a) ρx(ωm,p)\rho_x(\omega_{m,p}), where those for different pp are overlaid. (b) ρx(ωm,p)\rho_x(\omega_{m,p}) for the isolated modes with different ll. (c) Shift of the response ρx(ωm,p)\rho_x(\omega_{m,p}) for different chromaticities.

With the space-charge force at Np=7.5×1012N_p=7.5\times10^{12} (∣Δνsc/νz∣=27\lvert\Delta\nu_{sc}/\nu_z\rvert=27), instability occurs at −νR=νl+p−ξ/ηp-\nu_R=\nu_l+p-\xi/\eta_p, where νl\nu_l is the tune of an isolated mode related to the ll-th synchro-betatron mode. The growth rate decreases to 52%, 49%, 46% and 44% for l=1,2,3l=1,2,3 and 4 as shown in Figure 12(b) and Figure 13. The instability does not occur for l<0l<0 due to Landau damping. The growth of other bunch intensities NpN_p is obtained from the reduction factor seen in Figure 13.

The formulae are extended to the multi-bunch instability with HH bunches stored with equal spacing (C/HC/H),

(ν−νm)=−iHNprpβx4πγω0∑pZ(ωmμp)gm,L∗(κmμp−χ)gm,R(κmμp−χ)(80)(\nu-\nu_m)=-\frac{iHN_pr_p\beta_x}{4\pi\gamma}\omega_0\sum_p Z(\omega_{m\mu p})g_{m,L}^{*}(\kappa_{m\mu p}-\chi)g_{m,R}(\kappa_{m\mu p}-\chi) \tag*{(80)}

where μ=0,1,…,H−1\mu=0,1,\ldots,H-1 is the mode number of the coupled-bunch instability [10, 15] and ωmμp=(νm+μ+pH)ω0\omega_{m\mu p}=(\nu_m+\mu+pH)\omega_0 (κmμp=ωmμpσz/(cβ)\kappa_{m\mu p}=\omega_{m\mu p}\sigma_z/(c\beta)).

Conclusions

Understanding the oscillation modes amounts to representing the transformation of the system by a matrix and solving an eigenvalue problem. The beam distribution is represented by a vector of length 40×41=164040\times41=1640, and the transformation of the beam distribution is represented by a 1640×16401640\times1640 matrix. Solving this eigenvalue problem yields the oscillation modes of the bunch; the beam oscillation is expressed as a linear combination of the eigen-oscillation modes, and the time evolution and the linear response are analyzed. Solving the eigenvalue problem relies on numerical methods, but once the eigenvalues and eigenvectors are obtained, their treatment is standard linear algebra.

Oscillation modes of the beam were discussed for cases in which particles undergo free synchro-betatron oscillations, and for particles correlated by the space-charge forces. Eigen-oscillation states were found for both cases, considering the modulation due to chromaticity in the synchro-betatron modes. Without space-charge forces, eigen-oscillations exist at the betatron oscillation frequency and its sideband frequencies νx+lνz\nu_x+l\nu_z arising from periodic azimuthal displacements (ll) due to synchrotron oscillations, and are degenerate with respect to the radial direction in synchrotron phase space.

Space-charge forces lift this degeneracy, causing the tune to depend not only on ll but also on JJ. While a group of modes with continuously distributed tunes over JJ appears, for l≥0l\geq0, modes with discrete tunes, separated and isolated from the continuous tune region, emerge.

These tune distributions have a significant impact on the following phenomena in high-intensity proton rings, depending on the presence or absence of the space-charge force:

  • Decoherence and recoherence (echo) when a bunch is uniformly displaced as an initial condition,

  • Beam transfer function when forced oscillation is applied to a bunch,

  • Beam instability caused by a resonator impedance with a high quality factor Q≳10Q \gtrsim10.

The initial state of a kicked bunch contains a mixture of various eigenmodes when chromaticity is nonzero. By representing the initial state as a linear combination of individual eigenmodes and propagating each mode with its own eigenfrequency, decoherence and recoherence can be discussed. The linear coefficients are dominated by isolated modes, and as the chromaticity ∣χ∣|\chi| increases, isolated modes of higher order ll are included more. When forced oscillation is applied, modes corresponding to that frequency are excited. Only isolated modes are excited; continuous modes are not excited due to Landau damping. That is, l<0l < 0 modes are not excited. Excitation depends on integer multiples of the revolution frequency added to the tune of the forced oscillation. As the chromaticity χ\chi (or ξ\xi) is changed, the beam response does not change when the frequency is chosen so as to satisfy ν=νl+p−ξ/ηp\nu= \nu_l+p-\xi/\eta_p. The strength of the response depends on the space-charge force. The response is several times higher than that without the space-charge force. Beam instability is discussed based on the space-charge oscillation modes, namely, we assume that the oscillation modes are determined by the space-charge force, but are not affected by the impedance. The impedance with a low frequency ωRσz/(cβ)∼1\omega_R\sigma_z/(c\beta) \sim1 and a high quality factor Q≳10Q \gtrsim10 being studied in CSNS and J-PARC meets this condition. Instability growth and the beam-transfer function are expressed by the same formula. Only isolated modes with l≥1l \geq1 are excited by the impedance, but continuous modes distributed over ν<νx\nu< \nu_x are not excited due to Landau damping. The growth rate depends on the space-charge force. The growth for ∣Δsc/νz∣≥10|\Delta_{sc}/\nu_z| \geq10 is several times faster than that without the space-charge force.

Acknowledgements

This work is jointly supported by the Guangdong Special Support Program (Project No. 2024TQ08A570) and Guangdong Basic and Applied Basic Research Foundation (Project No. 2021B1515120021).

References

  1. [1]M. Blaskiewicz, Fast head-tail instability with space charge, Phys. Rev. ST-AB 1, 044201 (1998).DOI
  2. [2]A. Burov, Head-tail modes for strong space charge, Phys. Rev. ST Accel. Beams 12, 044202 (2009); 12, 109901 (2009) [erratum].arxiv.org/abs/0812.3914
  3. [3]V. Kormilov and O. Boine-Frankenheim, Head-tail instability and Landau damping in bunches with space charge, Phys. Rev. ST-Accel and Beams 13, 114201 (2010).
  4. [4]A. Macridin, A. Burov, E. Stern, J. Amundson, P. Spentzouris, Simulation of transverse modes with their intrinsic Landau damping for bunched beams in the presence of space charge, Phys. Rev. ST-Accel and Beams 18, 074401 (2015).DOI
  5. [5]Y. H. Chin, A. W. Chao, and M. M. Blaskiewicz, Two particle model for studying the effects of space-charge force on strong head-tail instabilities, Phys. Rev. Accel. Beams 19, 014201 (2016).DOI
  6. [6]X. Buffat, E. Metral, E. Gottlob, C. Hogh, A. Oeftiger, T. Pieloni, Description of beam instabilities in synchrotrons with wakefields and space charge forces using the circulant matrix model, Phys. Rev. Accel. Beams 24,060101 (2021).DOI
  7. [7]N. Yoshimura, T. Toyama, Y. Shobuda, Space charge effects on the intrabunch motion under large chromaticity at the main ring in the J-PARC, Phys. Rev. Accel. Beams 28, 101003 (2025).
  8. [8]R. Singh, O. Boine-Frankenheim, O. Chorny, P. Forck, R. Haselt, W. Kaufmann, P. Kowina, K. Lang and T. Weiland, Interpretation of transverse tune spectra in a heavy-ion synchrotron at high intensities, Phys. Rev. Accel. Beams 16, 034201 (2013).DOI
  9. [9]A. Burov, Coupled-beam and coupled-bunch instabilities, Phys. Rev. Accel. Beams 21, 114401 (2018).arxiv.org/abs/1606.07430
  10. [10]K. Ohmi, L. Huang, H. Liu and L. Rao, Space charge effects on coupled-bunch instability in high-intensity proton rings, Phys. Rev. Accel. Beams 28, 070101 (2025).
  11. [11]K. Oide, A mechanism of longitudinal single-bunch instability in storage rings, Part. Accel. 51, 43(1995), https://inspirehep.net/files/c68532571e82719dad511ab5f1e4ee76.
  12. [12]K. Ohmi, N.Kuroo, K.Oide, D.Zhou and F.Zimmermann, Coherent beam-beam instability in collisions with a large crossing angle, Phys. Rev. Lett. 119, 134801 (2017). K. Ohmi, Y. Zhang, C. Lin, Beam-beam mode coupling in collision with a crossing angle, Phys. Rev. Accel. Beams 26, 111001 (2023).
  13. [13]Liangsheng Huang, Sheng Wang, Shouyuan Xu, Xinguang Liu, Mingyang Huang, Renhong Liu, Biao Tan, Hanyang Liu, Li Rao, Yanliang Han, Source of instability in the rapid cycling synchrotron of the China Spallation Neutron Source, Eur. Phys. J. Plus 140, 71 (2025).DOI
  14. [14]Y. Shobuda, Y.H. Chin, P.K. Saha, H. Hotchi, H. Harada, Y. Irie, F. Tamura, N. Tani, T. Toyama, Y. Watanabe, M. Yamamoto, Theoretical elucidation of space charge effects on the coupled-bunch instability at the 3 GeV rapid cycling synchrotron at the Japan Proton Accelerator Research Complex, Prog. Theor. Exp. Phys. 2017, 013G01 (2017).DOI
  15. [15]A. Chao, Physics of Collective Beam Instabilities in High Energy Accelerators, Wiley-Interscience Publication, New York, 1993), and references therein.

Paper details

vlnc:W_3xp7x7ct
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC0-1.0
Usage

views · 5

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