Study on multi-fold bunch splitting in a high-intensity

medium-energy proton synchrotron

Linhao Zhang1,2,3, Min Chen2, and Jingyu Tang1,2,3,*
1
Institute of High Energy Physics, CAS, Yuquan Road 19B, Beijing 100049, China
2
University of Chinese Academy of Sciences, CAS, Yuquan Road 19A, Beijing 100049, China
3
Spallation Neutron Source Science Center, Zhongziyuan Road NO. 1, Dongguan 523803,
China

Bunch splitting is an RF manipulation method of changing the bunch structure, bunch

numbers and bunch intensity in the high-intensity synchrotrons that serve as the injector for a
particle collider. An efficient way to realize bunch splitting is to use the combination of
different harmonic RF systems, such as the two-fold bunch splitting of a bunch with a
combination of fundamental harmonic and doubled harmonic RF systems. The two-fold
bunch splitting and three-fold bunch splitting methods have been experimentally verified and
successfully applied to the LHC/PS. In this paper, a generalized multi-fold bunch splitting
method is given. The five-fold bunch splitting method using specially designed
multi-harmonic RF systems was studied and tentatively applied to the medium-stage
synchrotron (MSS), the third accelerator of the injector chain of the Super Proton-Proton
Collider (SPPC), to mitigate the pileup effects and collective instabilities of a single bunch in
the SPPC. The results show that the five-fold bunch splitting is feasible and both the bunch
population distribution and longitudinal emittance growth after the splitting are acceptable,
e.g., a few percent in the population deviation and less than 10% in the total emittance
growth.

I. INTRODUCTION
Beam manipulation in both the transverse and longitudinal phase space is a powerful
technique to attain desirable beam properties in synchrotrons, such as bunch length and
energy spread, emittance, bunch distribution and so on [1]. A variety of beam manipulation
methods in the longitudinal phase space, including phase-space stacking, debunching,
bunch-to-bucket transfer, bunch rotation, bunch merging or splitting, etc., have been applied
by employing special settings of RF systems in many accelerators [2, 3]. In particular, bunch
splitting is an RF manipulation method to change the bunch characteristics, like bunch
numbers, bunch intensity, and the time structure of beam, in high-intensity synchrotrons
serving as the injector of a particle collider.
In intense proton synchrotrons, two methods of splitting bunches have been utilized to
obtain the desirable bunch properties [3, 4]. One conventional method for changing the bunch
structure is to debunch the beam by turning off the RF voltage at the initial frequency, then to
let the beam circulate without longitudinal RF focusing, and finally to rebunch the beam with
another RF frequency. The other one is to adopt the combination of different harmonic RF

*
Corresponding author.
[email protected]
systems, e.g., the two-fold splitting of a bunch with a combination of fundamental harmonic
and doubled harmonic RF systems, which can minimize the longitudinal emittance growth
and beam loss as compared with the debunching method. The bunch splitting method has
been successfully applied to the LHC/PS for double splitting and triple splitting of one bunch
[5, 6]. It has also been proposed and studied through the simulation in the JLEIC (Jefferson
Lab Electron-Ion Collider) [7].
SPPC, as the second stage of the CEPC-SPPC (Circular Electron Positron Collider & Super
Proton-Proton Collider) project, aims at exploring new physics beyond the Standard Model.
The design luminosity at SPPC reaches 11035 cm-2s-1 with the nominal bunch spacing of 25
ns, which will produce a large number of events per bunch crossing and thus impose a severe
challenge on the detector trigger system and data analysis methods. To mitigate the pileup
effects and the collective instabilities of a single bunch at SPPC, the operation mode of
shorter bunch spacing, like 5 ns, was proposed at the CEPC Conceptual Design Report [8].
Likewise, the possibility of shorter bunch spacing has also been considered in the FCC-hh
(Future Circular Collider for hadron-hadron collisions) project [9]. Therefore, it is inevitable
and crucial to investigate the five-fold bunch splitting method to alter the bunch separation
from 25 ns to 5 ns. This manipulation will be conducted in the Medium-Stage Synchrotron
(MSS), the third accelerator of the SPPC injector chain.
The paper is organized as follows. The approaches of bunch splitting and the general
multi-harmonic RF bunch splitting method are described in Sec. II. In Sec. III, the simulation
method of bunch splitting and the application of the five-fold bunch splitting method to the
MSS are presented. Conclusions and discussion are given in Sec. IV.

II. BUNCH SPLITTING METHODS

There are two different bunch splitting methods used in synchrotrons: one is the
debunching and rebunching method; the other is the multi-harmonic RF bunch splitting
method. Both methods will have to be implemented under the stationary condition where the
beam energy and magnetic field are kept constant to achieve the uniformity and symmetry of
the bunch population after the splitting and relieve longitudinal emittance growth even the
beam loss. The details of the two bunch splitting techniques are described below.

A. Debunching and rebunching method

Debunching and rebunching is the most routine bunch manipulation method to change the
number of bunches and the bunch spacing [10]. Rebunching is the reverse process of
debunching. The procedures of this method are as follows: First, the RF voltage with the
original harmonic frequency is adiabatically lowered to a level where the momentum spread
is relatively low, then the RF system is turned off. Next, due to the turning-off of longitudinal
focusing, particles with different momenta drift with different revolution frequencies, then the
beam will fill up the whole ring and become the so-called coasting beam. Finally, a new RF
system with the desirable harmonic frequency is turned on, and then the RF voltage is
adiabatically increased to an appropriate value, during which the beam progressively receives
an azimuthal modulation of density and is finally fully bunched in the new harmonic RF
system.
To minimize the longitudinal emittance growth, the processes of debunching and
rebunching have to be adiabatic or quasi-adiabatic. Even so, the increase of emittance is
unavoidable since: (1) the RF voltage cannot be lowered down to a tiny amplitude, as the
strong beam loading effect may occur; (2) during the drifting when the beam is left
uncontrolled in the longitudinal plane, the space charge may introduce nonlinear effects and
instabilities; (3) the nonlinear focusing field of the new RF system that is added suddenly will
render the filamentation of the bunch in the phase space; (4) the coasting beam has a very
small momentum spread which leads to the possible longitudinal microwave instability. In
addition to the above drawbacks, as all the buckets are filled with bunches, the rebunched
beam may not leave a sufficient time gap for the extraction kickers, which will result in
intolerable beam loss during the extraction in high-intensity synchrotrons. Therefore, an
alternative bunch splitting approach was proposed [4], which will be summarized in Sec. II.B.

B. Multi-harmonic RF bunch splitting method

The main differences of the multi-harmonic RF bunch splitting method from the
debunching-rebunching method lie in: (1) during the splitting, multiple harmonic frequencies
coexist in the RF systems rather than appear successively; (2) there is not a phase of the full
missing of RF focusing; (3) the gaps between different bunch trains can be preserved. These
make the multi-harmonic RF bunch splitting method more suitable for bunch splitting in the
injector chain for a hadron collider. In the following, as the most straightforward example, the
two-fold bunch splitting that means one bunch splits into two bunches is introduced to
illustrate how this method works, and then more general multi-harmonic RF bunch splitting
will be discussed.

1. Two-fold bunch splitting with dual RF systems

Figure 1 illustrates the process of two-fold bunch splitting. First, one bunch is bound in the
stationary single-harmonic RF bucket with a harmonic number h0, see Fig. 1(a). Next, the
second-harmonic RF system with a doubled frequency or harmonic number 2h0 is turned on
to stretch the bunch, see Fig. 2(b). With the increase of the second-harmonic RF voltage, the
bucket center gradually becomes the unstable fixed point from the stable one, and two new
stable fixed points emerge on both sides of the bucket center. Then the bunch gets invaginated
in the bucket center, and two new bunches are formed, encircling the two new stable fixed
points, see Fig. 1(c)(d).

(a) (b)
 (c) (d)
FIG. 1. The schematics of the two-fold bunch splitting process.
To perform ideal bunch splitting, the phase difference between the two RF systems should
be , and the RF voltages should be programmed to follow the required pattern, which is
shown in Fig. 2. One can see that in the first half of the process, the gradual increase of the
doubled-harmonic RF voltage is to elongate the bunch, and in the second half, the gradual
decrease of the fundamental harmonic RF voltage helps divide the bunch into two smaller
ones.

FIG. 2. The RF program during the process of two-fold bunch splitting. Blue solid line:
fundamental harmonic RF voltage; orange dashed line: doubled-harmonic RF voltage.

2. Multi-bunch splitting with multiple harmonic RF systems

Based on the two-fold bunch splitting experience described above and the three-fold bunch
splitting experience in the LHC/PS, here we try to give the general method of one bunch
splitting into multiple bunches using multiple harmonic RF systems. The bunch splitting
method can be described as follows.
(1) The required number of harmonic RF systems depends on how many bunches one
bunch is to be split into. For instance, for one bunch into three, three simultaneous RF
systems with harmonic numbers h0, 2h0, and 3h0 are needed. It is worth noting that the
presence of the RF system with 2h0 is crucial, which assists in elongating the bunch before
splitting and producing three more-or-less uniformly populated sub-bunches.
(2) To generate equal bunch splitting, the relative RF phases among the required different
harmonic systems are definite. For bunch splitting in a synchrotron working below the
transition energy, the phases for RF systems with even multiples of h0 must be , whereas
those for RF systems with odd multiples of h0 must be zero. The opposite is true for the case
above transition energy.
(3) The voltages for different harmonic RF systems play critical roles during the bunch
splitting process. It is impossible to derive analytical solutions to the RF voltage programs of
different harmonic RF systems. However, one can choose the voltages at some pivotal
moments and finally determine them based on the numerical simulations. Therefore, the RF
voltage programs are not unique and are usually determined by trials. However, the
time-dependent trends of the voltages for different harmonic RF systems are regular. The RF
voltage with harmonic number h0 first keeps constant and then needs to decrease to zero
gradually. The voltages for the other harmonics should be increased progressively from zero
in sequential order of the harmonics and reach their respective peaks successively. Then the
voltages for the intermediate harmonics are decreased to zero in succession. Only the voltage
with the highest harmonic remains in the last period of the splitting.
It is worth pointing out that the above rule is more suitable for the cases where the number
of sub-bunches is a prime number. If the number of sub-bunches is a composite number, one
can manipulate the fundamental bunch splitting more than once. For example, if one wants to
split one bunch into four bunches, one can simply implement two-fold bunch splitting twice.

3. Key points for the RF voltage design

As partially mentioned above, to generate a good multi-fold bunch splitting with uniform
particle partition and small emittance growth, the most essential and challenging task is to
design the voltage patterns for the multiple harmonic RF systems. Besides the time-dependent
trends for the multiple harmonic RF voltages mentioned above, another two key points also
contribute to obtaining a better bunch splitting.
The first one is that the designed RF voltage scheme meets the adiabatic or quasi-adiabatic
condition, which means the change of RF voltages must be sufficiently slow so that the
redistribution of particles within the bucket is executed with an emittance growth as small as
possible. The condition for adiabatic synchrotron motion can be evaluated with the
adiabaticity coefficient ad as [2]:

1 d s 1 dTs
 ad = = 1. (1)
 2
s
dt 2 dt

Here, s and Ts are the angular frequency and period of the synchrotron oscillation,
respectively. Typically, the process is considered adiabatic when ad < 0.05. This adiabatic
condition is necessary to avoid the significant emittance growth by the filamentation effect of
the particle distribution in the presence of a strong nonlinear RF field.
Another key point is to make the bunch stretched as far as possible before splitting the
bunch or the intermediate bunch. Here, a general condition to flatten the bunch can be given
based on the experience from the one in a dual-harmonic RF system [11]. A generalized RF
voltage V() seen by the beam at a certain moment, which is the summation of different
harmonic RF voltages, can be expressed as:
n
V ( ) = V0  ki sin ( i + i ) (2)
i =1

where n is the number of RF harmonics,  is the phase in the fundamental RF, i is the phase
difference of the i-th harmonic RF to the fundamental RF and thus 1=0, V0 is the voltage of
the fundamental RF at the initial moment, ki is the voltage ratio of the i-th harmonic RF to the
fundamental RF at the initial moment. The corresponding potential well function can be
expressed by:

 V ( ) − V ( ) d
1 
U ( ) = s
s
V0
, (3)
n n

cos ( i + i ) − cos ( i + i ) − ( − s )  ki sin ( is + i )

ki
= s
i =1 i i =1

where s is the synchronous phase. To maximize the bunch length and obtain a flattened
potential well, the derivatives of V() from the first order to the n-th order should vanish at
the center of the bunch:
V ( )
=0, (4)
  =s

……
 nV ( )
= 0. (5)
 n  =s

Then, one can obtain odd=0 when i is odd, while even=0 when i is even, and different ki
values for all the harmonics. The highest order of the derivative is determined by the number
of the required equations to solve all the ki values. Figure 3 shows the flattened potential wells
obtained by using Eqs. (3)-(5) for n=2 (k2=0.5), n=3 (k2=0.8, k3=0.2), and n=5 (k2=8/7,
k3=9/14, k4=4/21 and k5=1/42). One can see that with more RF harmonics, one can obtain
relatively more flattened or wider potential well and thus longer bunch. It should be pointed
out that the above conditions are necessary to ensure no emergence of small inner buckets
within the original outer bucket. In practice, to generate the desired particle partition in the
higher-fold bunch splitting, one can adjust different k values to make the initial bunch even
more elongated than in the case of lower-fold bunch splitting.
2.5

2.0 n 2

1.5 n 3
n 5
1.0
U

0.5

0.0

0.5
3 2 1 0 1 2 3
rad

FIG. 3. The flattened potential well for multi-harmonic RF systems with the harmonics n=2
(red), n=3 (green) and n=5 (orange).

III. APPLICATION OF THE FIVE-FOLD BUNCH SPLITTING METHOD TO

THE MSS
 The nominal bunch spacing in the SPPC baseline design is 25 ns. There are 10080 bunches
in total for one of the two collider rings, each containing 1.51011 particles, to achieve the
nominal luminosity of 11035 cm-2s-1. The instant luminosity could be even higher when some
luminosity leveling methods are applied [12]. This will pose severe challenges on the detector
design and data analysis methods that will allow us to record the data first and extract the
interesting physics events afterwards. A natural idea to decrease the number of events per
bunch collision is to reduce the bunch intensity by splitting one bunch into several bunches,
which also means the reduction of bunch spacing. Although the instant luminosity is reduced
with a smaller bunch population or bunch spacing and with the same total protons, the bunch
spacing of 5 ns is still considered a good compromise. At SPPC, the solution to change the
bunch spacing from the nominal 25 ns to the optional 5 ns is to perform the five-fold bunch
splitting in the MSS, the third-stage accelerator of the SPPC injector chain. Likewise, the
possibility of 5 ns bunch spacing has also been considered in the FCC-hh, where the
formation of the bunch spacing of 5 ns will take place either by debunching in the LHC/PS
and rebunching in the LHC/SPS or by building a new injector chain [9, 13].

A. Introduction to the MSS

The Medium-Stage Synchrotron (MSS) is a high beam-power synchrotron, and the
debunching-rebunching method to change the bunch spacing does not work here. A special
lattice with a negative momentum compaction factor of -7.710-4 has been designed to
accelerate proton beams received from its injector, a proton rapid cycling synchrotron
(p-RCS), from 10 GeV to 180 GeV with a repetition rate of 0.5 Hz. Then the beams are
extracted into the downstream accelerator, a Super Synchrotron (SS). The circumference of
the MSS is about 3.5 km, which can accommodate three p-RCS beam batches, each of which
contains 112 bunches separated by 25 ns. The SPPC nominal bunch spacing of 25 ns is
formed from the bunch structure in the p-RCS with an RF frequency close to 40 MHz. The
nominal bunch intensity, defined as the number of particles in one bunch, is 1.51011 in the
MSS when providing beams for the SPPC.
The basic longitudinal dynamics parameters at the MSS injection and extraction are given
in Table I, where the bunch-to-bucket transfer match with its injecting synchrotron p-RCS and
the receiving synchrotron SS, the space charge effects and so on are taken into account [14].
In addition to the requirements of longitudinal beam dynamics parameters, the following
constraints during the process of bunch splitting should be considered: (1) The duration of the
whole bunch splitting should be controlled within 0.2 s as short as possible since the
repetition period of the MSS is only 2 s. (2) The process of bunch splitting should be
implemented under the adiabatic or quasi-adiabatic condition, which requires the voltages of
different RF harmonic systems should not change too fast. (3) In the case of being capable of
achieving bunch splitting, a relatively lower total RF voltage is preferred due to the expensive
cost of RF systems. (4) After bunch splitting, the total longitudinal emittance should not be
increased too much, e.g., not exceeding two times as that before splitting.

TABLE I. The basic longitudinal dynamics parameters at the MSS injection and extraction.
Parameter Unit Injection Extraction
Proton energy [GeV] 10 180
 Ring circumference [m] 3478 3478
Repetition rate [Hz] 0.5 0.5
11
Number of particles per bunch 10 1.5 1.5
Number of bunches 336 336
Bunch spacing [ns] 25 25
14
Accumulated particles 10 0.50 0.50
Longitudinal emittance, 4σ [eVs] 0.5 1.0
Revolution frequency [kHz] 86.3 86.3
Circulating beam current [A] 0.70 0.70
Beam power [MW] 0.04 0.73
Momentum compaction factor, p 10-3 -0.77 -0.77
RF frequency [MHz] 40 40
Harmonic number 464 464
Total RF voltage [MV] 3 5.8
Synchrotron frequency [Hz] 1106.65 118.46
Bucket area [eVs] 2.37 42.77
Bucket half-height, ΔE/E [10 ] -3
6.80 7.43
Full bunch length, 4σz [ns] 8.25 2.75
Momentum spread, σδ 10 -3
1.76 0.64
Momentum filling factor 0.50 0.17
Bunching factor 0.22 0.07

B. Simulation method for the bunch splitting

1. Introduction to the simulation code -- BLonD

BLonD (Beam Longitudinal Dynamics) is a simulation code of longitudinal beam
dynamics in synchrotrons developed by CERN since 2014 [15]. The code was developed and
implemented based on the widely-used Python language, and it has been benchmarked and
compared with measurements, theory, or other codes, which enhance its robustness and
credibility [16]. It can simulate single-bunch or multi-bunch longitudinal beam dynamics
during the acceleration cycle with multiple RF systems, multiple RF stations in different
locations, the collective effects, and the implementation of low-level RF controls such as
phase noise, phase loop and feedback. It has been applied practically to LHC controlled
longitudinal emittance blow-up during the ramp, PS-to-SPS transfer and so on. In this study, it
is used to simulate the bunch splitting process.

2. Methodological considerations of different bunch-splitting schemes

Based on the principal analysis of bunch splitting with multiple harmonic RF systems, as
mentioned in Sec. II.B, reasonable RF voltage schemes that will stretch the bunch and split it
eventually into multiple bunches need to be carefully studied. For two-fold bunch splitting,
one can readily split one bunch into two bunches uniformly with the combination of
fundamental harmonic h0 and doubled-harmonic 2h0 RF systems, owing to the phase-space
symmetry. No complicated RF voltage operation is required except for meeting the adiabatic
condition. For three-fold bunch splitting, one can first give a preliminary voltage pattern of
the three RF systems with harmonic numbers h0, 2h0 and 3h0, respectively, then adjust
individual RF voltage according to the simulation results of the particle redistribution in the
three sub-bunches after splitting. For example, if the middle sub-bunch particles are more
than the average, one could lower the h0 voltage, which can help decrease the depth of the
potential well in the middle, or increase the 2h0 voltage, which can help deepen the potential
well on both sides. The method of adjusting the RF voltages is straightforward, and the
equally-redistributed sub-bunches can be eventually obtained. However, the RF voltage
scheme design for the five-fold bunch splitting becomes much more intricate, which will be
explained in detail as follows.
The first fundamental problem is what kinds of harmonic RF systems we indeed require to
produce a five-fold bunch splitting. As stated in Sec II.B.2, the RF systems with h0, 2h0, 3h0,
4h0 and 5h0 are essential to manipulate five-fold bunch splitting. Here, it is demonstrated why
the intermediate RF systems with 2h0, 3h0 and 4h0 are indispensable in addition to the original
h0 harmonic and the final 5h0 harmonic. In fact, bunch splitting can be regarded as a process
of stretching the bunch and then redistributing particles within the lengthened bunch to the
newly established high-harmonic RF buckets under the control of multi-harmonic RF voltages.
During this process, elongating the bunch by keeping a more-or-less uniform linear density is
a key point. The more sub-bunches one bunch will split into, the longer it must be stretched
before the splitting. For example, the original bunch needs to be lengthened at least 2/3 for
the triple-fold bunch splitting, whereas it will be at least 6/5 for the five-fold bunch splitting,
as shown in Fig. 4. Otherwise, probably no particles will enter the outermost RF buckets on
both sides. To lengthen the bunch that far, an RF harmonic component with its peaks at 3/5
is needed. Assume the required RF harmonic component has a harmonic number kh0, where k
is an integer, it must satisfy:

3 2
2  even integer, when k is odd , (6)
5 k

3 2
2  odd integer, when k is even . (7)
5 k

Here, “” in Eqs. (6)-(7) means the peak appears at 3/5 or on its right so that as many
particles as the required can enter the outermost stable islands. The results show that k=3 can
better fulfill this condition. Hence, the RF system with 3h0 is vital. Besides, the RF system
with 2h0 is also crucial to move the core particles outward. Otherwise, the particle number in
the final center bunch will be much greater than in the other bunches, and this can be better
understood with the potential well. The RF system with 4h0 is also required to balance the
particle numbers in the outermost bunches and their neighboring bunches. Therefore, all the
RF systems with h0, 2h0, 3h0, 4h0 and 5h0 are essential to accomplish the manipulation of the
five-fold bunch splitting with equal particle redistribution.
FIG. 4. Requirement for the minimum phase extension for the five-fold bunch splitting.
Partial particles within the original bunch must be stretched beyond 3/5 so that they can
enter the outermost RF buckets on both sides.
Another fundamental problem is the complexity of the particle redistribution with the
voltage modulations of multiple harmonic RF systems to achieve the five-fold bunch splitting.
For the sake of simplicity, we name the five sub-bunches after the splitting as B1, B2, B3, B4
and B5 from left to right in the phase space plot. Thanks to the symmetrical bunch property,
the particle number in B1 and B5 is naturally equal, so as for B2 and B4. Thus, one only
needs to consider the equality of B1, B2 and B3. For instance, if the particle numbers are too
small in B1 and B2 but too large in B3, the increase of the 2h0 voltage can help reduce the
particle number in B3 and increase those in B2 and B1, and then the other voltages can be
adjusted to balance between B1 and B2. Unfortunately, there is no unique and accurate
method to generate the RF voltage programs of different harmonics. One has to optimize the
voltage programs of all the five harmonics based on the simulation results.

3. Initial RF voltage patterns

To give the initial RF voltage patterns of the five harmonics for the five-fold bunch
splitting, which are the input parameters for simulations with the BLonD code, one can design
the voltage curves at some pivotal moments based on the shape of the potential well.
Assuming the process of bunch splitting is adiabatic or quasi-adiabatic, the change in the
beam emittance shape always keeps pace with the change of the potential well. Hence, the
shape of the potential well determines the distribution of the beam in the longitudinal phase
space. One can optimize the voltage ratios (a set of k values) of different harmonics to obtain
the desired potential well based on Eq. (3). Figure 5 shows the potential wells for different
sets of k values. One can see that the potential well develops from the initial one (in green)
that corresponds to the single harmonic first to the flattened one (in red), then further
stretched along with the sequential formation of three and four small shallow wells (in blue
and orange), finally into five small shallow wells (in purple) with different sets of k values in
the different moments. During this process, the five harmonic RF systems with h0, 2h0, 3h0,
4h0 and 5h0 successively play predominant roles, which makes the potential well wider and
wider and redistributes the particles along with the phase continuously. Therefore, these
different sets of k values can be used as the original RF voltage schemes of the five-fold
bunch splitting, with the initial voltage of the fundamental harmonic determined by the
momentum filling factor.
 2.5
k1 1,k2 k3 k4 k5 0
2.0 k1 1,k2
8
,k3
9
,k4
4
,k5
1
7 14 21 42

1.5 k1 0.52 ,k2 0.78 ,k3 0.82 ,k4 0.28 ,k5 0.09
k1 0.13 ,k2 0.28 ,k3 0.37 ,k4 0.42 ,k5 0.24
1.0 k1 0.06 ,k2 0.13 ,k3 0.19 ,k4 0.24 ,k5 0.38

U
0.5
0.0

0.5
3 2 1 0 1 2 3
rad

FIG. 5. The shapes of the potential well in the different moments with different combinations
of harmonic voltages for the five-fold bunch splitting.

C. SIMULATION RESULTS

1. Five-fold bunch splitting at the extraction energy

First, we consider the possibility of realizing the five-fold bunch splitting at the extraction
energy of 180 GeV. This is natural as one may want to keep the original bunch spacing of 25
ns in the whole acceleration process. The multiple harmonic RF systems with the harmonic
number h0 (40 MHz), 2h0 (80 MHz), 3h0 (120 MHz), 4h0 (160 MHz) and 5h0 (200MHz) are
utilized to achieve the goal. The corresponding RF phases with respect to the bunch center are
0, , 0,  and 0. In the simulations, 50000 macro-particles representing the real particles of
1.51011 are tracked for about 17400 turns or 0.2 s. The initial bunch follows the
two-dimensional Gaussian distribution with the RMS bunch length of 0.688 ns and the RMS
absolute energy spread of 0.115 GeV. The intensity-related effects, such as the space charge
effects and collective instabilities, were not taken into account in the simulations because they
do not significantly affect the implementation of the bunch splitting technique but mainly
influence the inner particle redistribution.
With the RF voltage design method in Sec. III.B.3, the initial RF voltage patterns were
given at some pivotal moments with the initial voltage of 5.8 MV for the fundamental
harmonic RF, as shown in Fig. 6 (a). The final RF voltage of 1.2 MV with 5h0 was designed
to achieve the perfect match on the longitudinal phase space with the downstream SS.
Elaborate optimizations for the RF voltage patterns based on the multi-particle simulations
were conducted to obtain more homogeneous bunch populations after the splitting. The
optimized RF voltage patterns are presented in Fig. 6 (b). There is no significant difference
between the initial RF voltage pattern and the final one after optimization, except that two
other moments were added to the optimized one in the fine-tuning effort to obtain better
particle redistribution in the final five sub-bunches. This demonstrates that the method of
designing the RF voltage patterns based on the potential well is feasible and effective.
 (a) (b)
FIG. 6. The initial RF voltage patterns in (a) and the final one after optimization in (b) for the
five-fold bunch splitting at the extraction energy.
Figure 7 represents the evolution process of one bunch splitting into five sub-bunches in
the phase space with the optimized RF voltage patterns. One can clearly observe the processes
of the bunch stretching and splitting. Actually, the modulations of the different RF harmonic
voltages can be divided into two parts: in the first half, the RF systems with h0, 2h0 and 3h0
predominate, which is similar to triple splitting, and the original bunch evolves into three
bunches, see Fig. 7(a)(b)(c); in the second half, with the decrease of the RF voltages with h0,
2h0 and 3h0 and the increase of the RF voltages with 4h0 and 5h0, the middle bunch is further
lengthened and performs the triple splitting again, and gradually develops into three new
sub-bunches, while the other two bunches move outward on the two sides and become the
two outermost sub-bunches, see Fig. 7(d)(e). Finally, five equal-spacing sub-bunches are
formed under the RF voltage drive with 5h0, see Fig. 7(f), which is assisted with 4h0 in the
process.
The original bunch containing 50000 macro-particles is ultimately divided into five
sub-bunches with populations of 9599, 10278, 10170, 10090 and 9863 from B1 to B5 (or
from left to right). The maximum population deviation is about 4.0%, which can be
acceptable. However, the longitudinal emittance of the five sub-bunches containing 99%
macro-particles is 1.333 eVs, 1.215 eVs, 1.169 eVs, 1.225 eVs and 1.342 eVs, respectively,
which is much larger than one-fifth of the original longitudinal emittance of 2.424 eVs. This
means that the total emittance growth is about 1.59 times. In principle, the longitudinal
emittance after the splitting would not blow up significantly under the adiabatic condition.
The main cause for the large emittance growth is that the synchrotron tune under the five
harmonic RF systems varies dramatically across a much-elongated bunch. Besides, it is
difficult to satisfy the adiabatic condition with a given period of 0.2 s and a relatively small
synchrotron tune at the extraction energy—the consequent strong filamentation of the
emittance results in the large emittance growth.
 (a) (b)

(c) (d)

(e) (f)
FIG. 7. The bunch distribution in the longitudinal phase space and their projections on the two
axes after different turns during the five-fold bunch splitting at the extraction energy. (a) 0th
turn; (b) 5400th turn; (c) 7400th turn; (d) 12600th turn; (e) 14400th turn; (f) 17400th turn.

2. Five-fold bunch splitting at the injection energy

As compared with the bunch splitting at the extraction energy, the same manipulation at the
injection energy has the advantage of a larger synchrotron tune that is almost 10 times higher
than that at the extraction energy and helps meet the adiabatic condition, although the whole
acceleration will be executed by the 5h0 harmonic RF system instead of the original
fundamental harmonic RF system. In this case, the kinetic energy is 10 GeV, the RMS bunch
length is 2.06 ns, and the RMS absolute energy spread is 0.019 GeV. Both the initial voltage
of the fundamental harmonic RF and the final voltage of the 5h0 harmonic RF are set to 3 MV.
The RF voltage patterns after optimization and the corresponding splitting results are
illustrated in Fig. 8 and Fig. 9, respectively. After splitting, the populations of the five
bunches are 10058, 10024, 9828, 10150 and 9940, respectively. The maximum relative
deviation in population is around 1.7%. The original longitudinal emittance is 1.194 eVs, and
the one after the splitting is 0.246 eVs, 0.267 eVs, 0.261 eVs, 0.268 eVs and 0.252 eVs for
the bunches from B1 to B5. The total emittance growth is less than 10%, which is much better
than that at the extraction energy.
It is worth pointing out that there are about 872 macro-particles beyond the RF buckets at
the end of the splitting, which account for 1.74%, due to the dramatic reduction of the RF
bucket area. These particles could be lost during the following acceleration. If the 2D
Gaussian distribution for the initial bunch is replaced by a truncated Gaussian of 3 times RMS
that almost reflects the usual cases in synchrotrons, so as in the p-RCS, then the number of the
potentially lost particles decreases significantly, which only accounts for 0.094%, as shown in
Fig. 10. Thus, it reveals that the potentially lost particles mainly come from the beam halo
particles of the initial distribution. Besides, the corresponding beam loss rate is about 0.01
W/m, far less than the uncontrolled beam loss standard of 1 W/m for hands-on maintenance
[17]. Therefore, the above scheme for the five-fold bunch splitting at the injection energy
looks fully acceptable.

FIG. 8. The RF voltage patterns after optimization for the five-fold bunch splitting at the
injection energy.
 (a) (b)

(c) (d)

(e) (f)
FIG. 9. The particle distribution in the longitudinal phase space and their projections on the
two axes after different turns during the five-fold bunch splitting at the injection energy. (a)
0th turn; (b) 5400th turn; (c) 8600th turn; (d) 10600th turn; (e) 13200th turn; (f) 17400th turn.
There are about 1.74% of macro-particles beyond the RF buckets at the end of the splitting,
which are shown in gray dots in (f).

(a) (b)
FIG. 10. The initial bunch distribution with a Gaussian truncated at 3 times RMS (a) and the
final bunch distribution after the five-fold bunch splitting (b). There are only about 0.094% of
macro-particles beyond the RF buckets at the end of the splitting.

3. Phenomena and result analysis

The manipulation of the five-fold bunch splitting using the multiple harmonic RF systems
can be achieved both at the injection energy and the extraction energy within the limited
duration of 0.2 s. However, the phenomena during the bunch splitting obviously become more
complicated under five harmonic RF systems than those with the two-fold or three-fold bunch
splitting. For instance, the bunch is slightly tilted at the 5400th turn during the bunch
lengthening process at the extraction energy but not at the injection energy, see Fig. 7(b) and
Fig. 9(b), mainly since the adiabatic condition is better satisfied at the injection energy than at
the extraction energy. Besides, the bunch centers of B2 and B4 after the splitting are hollow or
sparse for both at the injection energy and at the extraction energy, but more serious for the
latter, see Fig. 7(f) and Fig. 9(f). Here, taking the case of the extraction energy as an example,
some explanations are given for understanding the phenomena, which is based on the
synchrotron frequency distribution in the potential well.
Similar to the beam dynamics in dual harmonic RF systems [18], the Hamiltonian in
multiple harmonic RF systems can be given by:

Qs 0
H= P 2 + Qs 0U ( ) , (8)
2

h0  p h0 eV0 
with P = − and Qs 0 = . (9)
Qs 0 p0 2 2 E0

Here, P is the normalized momentum with  the slippage factor and p/p0 the fractional
momentum deviation; Qs0 is the synchrotron tune at zero amplitude for the fundamental
harmonic RF system with  the relative velocity and E0 the synchronous energy; U() is the
potential well function, as shown in Eq. (3). For a given trajectory with the maximum phase

angle ˆ in the longitudinal phase space, the corresponding Hamiltonian becomes:

H 0 = Qs 0U ˆ . ( ) (10)

Moreover, the corresponding action is given by:

ˆ ˆ
2 U (s ) − U ( )d .
1 1 1
J ( H0 ) =  Pd =  Pd =  (11)
2  − ˆ
 −ˆ

Thus, the corresponding synchrotron tune can be calculated numerically by:

dH 0
Qs = . (12)
dJ
Figure 11 shows the calculated potential well and the corresponding synchrotron tune at the
14400th turn at the extraction energy based on Eqs. (8)-(12). Considering the symmetry, we
can merely calculate the synchrotron tune between 0 and . One can see that the synchrotron
tune distribution has several segments, which mainly depends on the number of valleys of the
potential well. It is worth noting that the synchrotron tunes at the centers of B2 and B4 at the
14400th turn collapse to zero, whereas the tunes are much larger on both sides of the centers.
Therefore, during the formation of B2 and B4, the particles after coming out from the middle
bunch will be more prone to oscillate around the centers of B2 and B4 and form a spiral
distribution rather than a more compact distribution with the full filamentation effect in a
gradually growing bucket. Accordingly, the interiors of the B2 and B4 are somewhat sparse,
as shown in Fig. 7(f). The situation gets significantly better at the injection energy, as shown
in Fig. 9(f), because the adiabatic condition can be better satisfied. It will certainly result in
smaller emittance growth.

(a) (b)
FIG. 11. The potential well in (a) and the corresponding synchrotron tune in (b) at the
14400th turn at the extraction energy. One can see that the synchrotron tune at the center of
B4 or about 1.1 rad collapses to zero. There are some spikes from numerical calculation errors
in the synchrotron tune plot.
Table II summarizes the main properties of the five-fold bunch splitting at the injection
energy and the extraction energy. The problem of the longitudinal emittance growth after
splitting is considerably mitigated at the injection energy as compared with that at the
extraction energy. The main reason is that the synchrotron frequency at the injection is much
larger, thanks to the larger slippage factor at the lower energy, so that the adiabatic condition
can be better satisfied. Besides, the uniformity of the bunch population after the splitting can
be well under control at the injection energy. Although there is some beam loss, it is totally
below the tolerable limit, especially when the bunched particles from the p-RCS meet the
truncated Gaussian distribution. Therefore, it is better to manipulate the five-fold bunch
splitting at the injection energy than at the extraction energy.
TABLE II. The comparison of the simulation results of the five-fold bunch splitting at the
injection energy and extraction energy.
Schemes Beam loss Max. deviation in Total longitudinal
rate the population emittance growth
Injection - Gaussian 1.74% 1.7% 8.4%
Inj. - truncated Gaussian 0.094% 1.4% 9.1%
Extraction - Gaussian 0 4.0% 159%

IV. Conclusions and discussion

In this paper, the general multi-fold bunch splitting technique with multi-harmonic RF
systems is summarized, and it is the first time to study the manipulation of five-fold bunch
splitting. This method involves the selection of the number of harmonic RF systems, the
determination of the relative phases of different harmonics, and the design and optimization
of each harmonic voltage that are the most important and challenging. The method of
designing the RF voltage patterns at some pivotal points based on the potential well under the
adiabatic condition is first applied, and the further optimization of the voltages is based on the
multi-particle simulations. Finally, the synchrotron tune distribution across the bucket in the
multi-harmonic RF systems has been numerically calculated to understand the complicated
phenomena related to the particle redistribution during the splitting.
The five-fold bunch splitting method using multi-harmonic RF systems has been
successfully applied to the MSS, the third stage of the SPPC injector chain, to obtain a bunch
spacing of 5 ns from the original 25 ns. The results show that the bunch population
redistribution after the splitting can be well controlled both at the injection energy and
extraction energy, but the control over the longitudinal emittance growth is significantly better
with the former. The main reason for the difference is that the adiabatic condition is much
better satisfied at the injection energy within a limited period of 0.2 s. With the bunch
population deviations of less than 1.7% and the total longitudinal emittance growth of less
than 10%, the scheme for the five-fold bunch splitting at the injection energy is considered
promising and acceptable. Certainly, the method can also be applied to other high-intensity
proton synchrotrons or other accelerators requiring multi-fold bunch splitting.
In the future, further studies are needed to make the five-fold bunch splitting method fully
applicable in the SPPC. The effects related to beam intensity, like the space charge effects, the
beam loading effects and the collective instabilities, should be considered, as they will
aggravate the emittance growth and change the uniformity of bunch redistribution and even
cause beam loss.
 ACKNOWLEDGMENTS
The authors would like to thank Alexandre Lasheen of CERN, one of the BLonD
developers, for beneficial help and guidance for the use of the BLonD code. The study is
supported by the National Natural Science Foundation of China (11575214, 12035017) and
CAS-IHEP Fund for PRC\US Collaboration in HEP.

REFERENCE
[1] Michiko G. Minty, and Frank Zimmermann, Beam Techniques--Beam Control and
Manipulation, in Proceedings of Lectures given at the US Particle Accelerator School, June
14-25, 1999; Report No. SLAC-R-621, April 2003. https://doi.org/10.2172/813024.
[2] S.Y. Lee, Accelerator Physics (World Scientific, Singapore, 2004), pp. 317-340.
[3] R. Garoby, RF gymnastics in synchrotrons, in Proceedings of CERN Accelerator School
2010—RF for accelerators, Ebeltoft, Denmark, 2010, arXiv:1112.3232.
[4] R. Garoby, Bunch merging and splitting techniques in the injectors for high energy hadron
colliders, in 17th International Conference on High Energy Accelerators (HEACC'98), Dubna,
Russia, 1998, pp.172-174; Report No. CERN-PS-98-048-RF,
http://cds.cern.ch/record/367499.
[5] R. Garoby, S. Hancock, and J.L. Vallet, Demonstration of bunch triple splitting in the
CERN PS, in 7th European Accelerator Conference, Vienna, Austria, 2000; Report No.
CERN-PS-2000-038-RF, http://cds.cern.ch/record/453506.
[6] R. Garoby, Multiple bunch-splitting in the PS: results and plans, Report No.
CERN-PS-2001-004-RF, Geneva, Switzerland, 2001, http://cds.cern.ch/record/488252.
[7] R. Gamage and T. Satogata, Bunch splitting simulations for the JLEIC ion collider ring, in
Proceedings of the 7th International Particle Accelerator Conference, Busan, Korea, 2016, pp.
2448-2450, https://doi.org/10.18429/JACoW-IPAC2016-WEPMW013.
[8] The CEPC-SPPC study group, CEPC Conceptual Design Report, Volume I-Accelerator,
Report No. IHEP-CEPC-DR-2018-01, IHEP-AC-2018-01, 2018, Chap. 8, arXiv:1809.00285.
[9] Abada, A., Abbrescia, M., AbdusSalam, S.S. et al., FCC-hh: The Hadron Collider. Eur.
Phys. J. Spec. Top. 228, 755–1107 (2019). https://doi.org/10.1140/epjst/e2019-900087-0.
[10] R. Garoby, PS Machine Development Report Debunching (h=20) and Rebunching (h=84)
at 26 GeV in the PS, Report No. PS/RF/Note 98-02 (MD), 1998.
http://cds.cern.ch/record/957830/files/cer-002624160.pdf.
[11] A. Hofmann and S. Myers, Beam dynamics in a double RF system, in Proceedings of
11th International Conference on High-Energy Accelerators, CERN, Geneva, Switzerland,
1980, pp.610-14; Report No. CERN-ISR-TH-RF /80-26, http://cds.cern.ch/record/879237.
[12] L.J. Wang and J.Y. Tang, “Luminosity optimization and leveling in the Super
Proton–Proton Collider”, Radiat Detect Technol Methods (2021).
https://doi.org/10.1007/s41605-020-00233-6.
[13] H. Bartosik, H. Damerau, and E. Shaposhnikova, Implications of 5 ns bunch spacing for
the FCC-hh injector chain, in Future Circular Collider Week 2017, Berlin, Germany, 2017,
https://indico.cern.ch/event/556692/contributions/2567943/attachments/1468210/2271031/spa
cing5nsFCC_final.pdf.
[14] L.H. Zhang, J.Y. Tang, Y. Hong, Y.K. Chen and L.J. Wang, Optimization of Design
Parameters for the SPPC Longitudinal Dynamics, arXiv:2101.10623.
[15] CERN Beam Longitudinal Dynamics code BLonD, http://blond.web.cern.ch.
[16] H. Timko, J. Esteban Müller, A. Lasheen, and D. Quartullo, Benchmarking the beam
longitudinal dynamics code BLonD, in Proceedings of the 7th International Particle
Accelerator Conference, Busan, Korea, 2016, pp.3094-3097,
https://doi.org/10.18429/JACoW-IPAC2016-WEPOY045.
[17] J.Y. Tang, Rapid cycling synchrotrons and accumulator rings for high-intensity hadron
beams, Reviews of Accelerator Science and Technology, Vol. 6, (2013),
https://doi.org/10.1142/S1793626813300077.
[18] J.Y. Liu, D.D. Caussan, M. Ellison, S.Y. LEE, D. LI, A. Riabko and L. Wang, Analytic
solution of particle motion in a double rf system, Part. Accel. 49 (1995) pp.221-251,
http://cds.cern.ch/record/1120213.