Generation of Electron Bunches at Low Repetition Rates Using a Beat-Frequency Technique

Even at a continuous wave facility such as CEBAF at Jefferson Lab, an electron beam with long time intervals (tens of ns) between individual bunches can be useful, for example to isolate sources of background via time of flight detection or to measure the energy of neutral particles that cannot be separated with a magnetic field. This paper describes a demonstrated method to quickly and easily deliver bunches with repetition rates of 20 to 100 MHz corresponding to time intervals between 10 to 50 ns (respectively). This is accomplished by changing the ON/OFF frequency of the RF-pulsed drive laser by a small amount ( Δ f/f < 20%), resulting in a bunch frequency equal to the beat frequency between the radio frequencies of the drive laser and the photoinjector chopper system.


INTRODUCTION
The Continuous Electron Beam Accelerator Facility (CEBAF) at the Thomas Jefferson National Accelerator Facility (Jefferson Lab) employs synchronous photoinjection, where RF-pulsed lasers are used to create an electron beam with RF structure directly at the photocathode [1]. Under normal operating conditions, three lasers, one for each experimental hall, emit ~ 40 ps (FWHM) optical pulses with 499 MHz pulse repetition rate corresponding to the third subharmonic of the CEBAF accelerating cavities. The light from these lasers is combined and directed at the same location on a GaAs photocathode within the 100 kV DC high voltage photogun to produce three independent electron pulse trains interleaved in time and each experimental halls receives an electron beam with 2 ns bunch spacing.
For some nuclear physics experiments, it is desirable to operate with larger bunch spacing. For example, the G 0 Forward Angle experiment [2] measured parity violating asymmetries in elastic electron-proton scattering. An electron beam with 32 ns bunch spacing was required to help distinguish elastically scattered protons from "slower" low energy background composed of pions and inelastically scattered protons. It took hours to install and remove the large modelocked laser used to produce the necessary time structure. In this paper, a new technique is described to quickly create a low repetition rate beam with large time intervals between bunches using the same lasers that normally operate at 499 MHz. This is accomplished by operating the drive laser at a pulse repetition rate only slightly different from the normal operating frequency. The experimental hall receives beam at the beat frequency between the drive laser and the photoinjector chopper. Repetition rates from 20 to 100 MHz were demonstrated, with bunch separation ranging from 50 to 10 ns. This beat frequency technique is considerably easier to implement compared to installing a completely new laser, and although it provides only a few microamperes of low duty factor beam, we expect it will become a valuable and widely used tool for nuclear physics experiments at CEBAF.  Table 1 shows a partial list of laser frequencies derived from these equations. Clearly, many laser frequencies can be chosen that provide bunch spacing > 10 ns, a time interval suitable for applications described below. Initial tests to demonstrate the feasibility of the method were performed using a portable RF signal generator (Hewlett-Packard Model 8662) in place of the stabilized RF master oscillator normally used to operate the photogun drive laser. Only integer frequencies were used because the RF source could not accommodate frequencies with run-on decimals. These frequencies are highlighted red in Table 1. The laser pulse frequency was varied between 399 and 479 MHz, to obtain bunch spacing from 10 and 50 ns, respectively.
In "beat frequency" mode, most of the extracted beam is dumped at the photoinjector chopper, as illustrated in Figure 1 and indicate that it should be possible to provide a few microAmperes beam current with high frequency contamination < 1 % for bunch spacing less than ~ 30ns. For longer bunch intervals, high frequency contamination becomes very large (see 50ns example).
Note, however, that contamination can be reduced at the expense of delivered beam current by reducing the width of the chopper slit.

LASER AND RF SYSTEM
Measurements were made using a new drive laser [4] that consists of a fiber-coupled, gain-switched diode seed laser and ErYb-doped fiber amplifier at 1.56 μm wavelength, followed by a periodically poled lithium niobate crystal to obtain useful light at 780 nm via frequency doubling (Figure 3, top). The new laser is ideally suited for implementing the beat-frequency technique because the pulse forming mechanism, gain-switching, does not depend on laser cavity length as does modelocking. Picosecond pulses can be obtained over a wide range of frequencies by merely changing the frequency of the RF that is applied to the seed laser.
As mentioned above, initial tests to demonstrate the feasibility of the method were performed using a portable RF signal generator. Although this RF source was adequate to demonstrate the feasibility of the beat-frequency method, significant timing jitter could be observed on the beam. Greatly improved beam quality was obtained using the frequency divider and mixer circuit shown in Figure 3 (bottom). The 499 MHz accelerator master oscillator (MO) signal was divided by integer N=16 using an emitter coupled logic prescaler to produce the desired low frequency bunch repetition rate. This signal was then re-mixed with the 499 MHz MO to produce the proper beat-frequency sideband which was then filtered, amplified and applied to the drive laser. This versatile approach was easy to implement and ensured phase-coherence between the atypical laser frequency and the chopper frequency at 499 MHz (after N cycles), as well as maintaining the long-term timing stability. In addition, employing a low-noise, programmable divider such as an Analog Devices AD9511 facilitates any bunch spacing in 2 ns increments, even frequencies with run-on decimal.
APPLICATIONS MØller polarimtery: a study of potential systematic errors Electron beam polarization measurements that rely on MØller scattering are performed at low current (< few uA) to avoid target heating and subsequent target depolarization, while nuclear physics experiments are typically conducted at significantly higher current.
This inconsistency between beam current conditions has been the source of concern for some nuclear physics experimenters who would rather measure polarization at the same beam current as required by the experiment, to avoid systematic errors that might arise, for example, due to possible current-dependent photocathode phenomena.
In addition, Hartmann et al. [5], have determined that beam polarization can vary across the temporal profile of a narrow electron bunch (few ps). Specifically, the leading edge of the pulse can have higher beam polarization compared to the trailing edge, because electrons depolarize as they diffuse to the surface of the GaAs crystal. This behavior has been observed at CEBAF, although under normal experimental conditions, nearly the entire electron bunch is delivered to the physics target and variations in polarization associated with different electron diffusion times are averaged out; experiments receive beam polarization representative of the average of all of the electrons within the bunch [6]. Problems can arise during polarimetry, however, when a narrow chopper slit is used to reduce beam current to an acceptable level [7]. A narrow chopper slit passes only a small portion of the electron bunch. Chopper-slit misalignment or operation with an incorrect laser phase results in delivery of electrons from the leading or trailing edge of the bunch. As a result, the experimenters measure beam polarization that is higher or lower than actual values associated with production running.
The beat frequency technique was employed to study the magnitude of the systematic errors described above. While using the beat frequency technique, extracted beam current from the photogun remains high and consistent with levels used during the experiment, but delivered beam current is low because only some of the extracted electron pulses pass the chopper. Moreover, the chopper slit remains open and the entire bunch is delivered to the polarimeter, eliminating the possibility that beam polarization can be sensitive to different electron diffusion times within the photocathode.
Polarization measurements were performed using the MØller polarimeter [7,8] at The beat frequency technique was employed to deliver beam with long electron bunch spacing for the GE n experiment [11] at experimental Hall A [12], to help experimenters calibrate the timing resolution of the neutron detector and scintillator planes of the Big Bite spectrometer. This sophisticated spectrometer has a timing resolution of ~ 670 ps (one sigma) and subtle contributions due to pathlength and timewalk corrections are difficult to discern using normal CW beam with 2 ns electron bunch spacing.
In addition, an electron beam striking a target or beam dump creates neutrons with a wide energy spectrum and it may take many nanoseconds for this neutron "background" to dissipate. Figure 5 shows the distribution of neutron arrival times at the detector with respect to the electron beam, indicating that the slowest neutrons arrive at the detector within ~ 20ns. This information allows the experimenters to apply neutron energy cuts while analyzing data. Such information could not have been obtained using normal CW beam with 2 ns electron bunch spacing.

Particle Identification at Experimental Hall C
A Cerenkov detector is a key component of the backward angle portion of the G 0 parityviolation experiment [13]. Standard CW electron beam with 2 ns bunch spacing cannot be used to determine Cerenkov counter efficiency for electrons and pions because signals created by these particles are separated by only ~ 3 ns. To assist the experimenters with particle identification, the G 0 toroidal magnetic spectrometer was used to measure particle momenta and the beat frequency technique was employed to produce electron bunch spacing of 32 ns to implement TOF detection to cleanly distinguish between pions and electrons using the relation m o = p(β -2 -1) 0.5 ( Figure 6). Knowledge of the particle identity coupled with the existence or non-existence of a corresponding signal in the Cerenkov counter was used to determine the counter's efficiency for electrons, a quantity necessary for understanding the systematic uncertainties in the experiment.

CONCLUSIONS
The beat-frequency technique provides a simple way to create an electron beam with long time intervals between bunches, a highly useful tool for nuclear physics experimenters at CEBAF. Already, this beam has been delivered to two experimental halls to study potential sources of polarimeter systematic error, to investigate sources of unwanted background, particle identification and for detector calibration. The versatile drive lasers and the sophisticated injector chopper system make this possible. One obvious drawback of the beat frequency technique is that most of the extracted beam gets dumped on the chopper aperture. The second drawback is that at high current, electron-bunch lengthening due to Coulomb repulsion introduces contamination from neighboring electron bunches, limiting the average beam current to a few microamperes. Despite these limitations, the method has proved to be a very useful tool in experimental data analysis and it is expected that future experiments at CEBAF will routinely use this additional feature of the polarized photoinjector.   Beam within the vertical lines will be delivered to the experimental hall. These plots illustrate how contamination at the laser repetition rate is introduced onto the beam, a