The demonstration laser requirements are set 1) by the desire to provide usefullight output so that demonstrations, tests, and optimizations can be performed,and 2) to begin to address the requirements of a system scaled up in power thatwould satisfy the ultimate shipboard needs. The primary achievement desiredfrom this effort is the increase in average power out of a free-electron laserby a factor of 100 over what has been achieved to date. This places demands onthe linac injector for high average current at high electron beam quality andon the optical cavity for good alignment and figure control with high averageflux at short wavelengths. Details of these requirements are set out below indiscussions under each WBS element. The top-level requirements are set outbelow with design goals (those parameters we are designing to) as well asminimum requirements (those values which must be met to carry out successfullythe tests and begin to address the scale-up issues).

Wavelength Range

The ultimate desire is to achieve high power in the 1 to 2 micron wavelengthrange. This will not be possible in the initial incarnation of the IR Demobecause of energy limitations of the accelerator. The initial design goal is 2to 6.5 microns but the system will be designed with the capability for upgradeto the shorter wavelength range. It may be possible to operate at thirdharmonic of the fundamental at substantially reduced power. The wider range isprovided to allow for investigation of atmospheric transmission over a widerrange and unknown uses which may arise. The good mirror technology availablehelps operation at the short-wavelength end. At the longer wavelengthsdiffraction becomes an issue, leading to higher optical losses which eventuallyprevent lasing at the long-wavelength end.


The design goal is in excess of 1 kW with the minimum set as 1 kW. This is setby the desire to demonstrate the ability of the technology to produce asubstantial laser power output. One kilowatt was recognized as the entry pointto high-power laser operation.


The ultimate system must be packaged in a configuration consistent withshipboard installation. The initial system has not been compacted to theextent ultimately required but key features which permit that compaction willbe shown: superconducting cavity operation and energy recovery. Typically thevolume of FELs is not large but the layout is awkward to retrofit in anexisting vessel because of the large aspect ratio.

Operating Environment

The initial system uses superconducting modules developed for CEBAF's linacswhich operate in a quiet tunnel isolated from environmental noise. Ultimatelythe system must operate reliably in a shipboard environment which vibrates andflexes and in which the direction of gravity varies. These will all requireengineering solutions which are different from existing sub-systems. A spaceframe design will be needed for the optical cavity and electron transportsystems.


The modules that make up a FEL are inherently transportable for the mostpart. Carrying that characteristic to a higher level should be accomplished onthe first system to be installed in a shipboard environment. Key modificationswill be needed to install the cryogenic systems, especially transfer lines, ina more palletized way. Other systems will benefit from thought aboutconnections to power, cooling, etc. in a more standardized bus than has beendone to date.

Optical Beam Quality

FELs intrinsically produce high optical beam quality. Typically they quitlasing before the beam quality degrades. The requirements are set by thedesire to focus the beam to a tight spot on a target and on a coarser scale tominimize transport loss through the optical transport train. The goal is setas 1.4 times the diffraction limit with better than two times diffraction limitas the minimum requirement. Provided care is taken in the transport optics tothe user laboratory, it is expected that the FEL output beam quality can bemaintained to the user with perhaps only defocus needing compensation as themirrors heat.


Bandwidth requirements are set by the desire to operate on absorption peaks(relatively broad) and by the desire to avoid chromaticity in any transmissiveoptics. The FEL should produce a near transform limited pulse, provided thecavity detuning is properly set and jitter in FEL gain and electron beam energyis small (these are precisely defined in the derived requirements set outbelow). Analysis indicates that above about 180 nm a bandwidth within 4 timesFourier will maintain the expected focal spot size within the 2 timesdiffraction-limited size. The impact of wider bandwidth is effectively similarto bad beam quality. That has been chosen as the minimum requirement.


The need for stability is set by the desire to perform reasonably controlleddemonstrations. This means that the power must be stable, the wavelength can'tmove around, the power must be on long enough to allow setup, alignment, andcalibration, and the beam must go where it is supposed to all the time. Thevalues are all judgment calls with no hard edges in performance at this stage,but failure to achieve some reasonable level of performance would put intoquestion our ability to scale the system up to an useful machine. The powerstability is set at better than 25% for exposure control. Eventually betterthan 10% is required. This is rather demanding because of the indirect controlknobs available for feedback into the laser. On a slow time scale the beamaverage current can be controlled. To set up the machine it is necessary thatthe beam stay on without tripping a hardware safety monitor. If this occurstoo often, operation will become too erratic to tune the laser. We have setone trip every 10 minutes as the maximum, with less than one trip every hourdesired. The bandwidth variation sets tight specifications on the energy andphase jitter of the linac. We have chosen 1% as the minimum variation allowed,with a desire for 0.2%. The pointing jitter is set by the desire to focus to agiven spot. The minimum value is set at 10 mrad by what is expected fromsimple engineering. Eventually better than 1 mrad control is desired; thiswill take active pointing control.

Beam Interception

Beam interception is set by the desire to minimize backgroundradiation and thus shielding requirements. Eventually a high-power machineshould have this minimized to make the system acceptable on a ship. In thepresent system the allowed value will be set by the facility wall thickness orother shielding preferably at less than 1 nA at energies above 25 MeV, whereneutron production is severe. The facility shielding is designed for a totalinterception of 10-3 of the beam. In the initial system we willpermit beam interception at chosen points that can be shielded, provided thepower loading and gas evolution are not problematic.


It is eventually desired to run the laser reliably for short periods of time.Unforeseen shutdowns are particularly undesirable and must be limited infrequency and duration. More than 1000 h between failures is desired. Theinitial operation will be limited by photocathode life. Experience at SLACsuggests it should be possible to get more than 40 h on a cathode at full beamcurrent before an overnight reactivation will be required. However, no data atthe average currents desired presently exists from a comparable system. Wehave chosen 2.5 h at full current as the minimum which would allow meaningfuldata to be obtained. It is expected that initial setup would occur at loweraverage currents and therefore have longer gun life. Only after the system wasset up and ready would the full-power run be performed.

Energy Recovery

Energy recovery is needed in the present system because the installed RF isinsufficient to generate the required output power unless recovery works closeto design values. In the scaled system it is crucial so that system size andenergy use is acceptable. We have set the minimum level as 70% becauseanything less than this value would be inconsistent with our currentunderstanding of how the system operates and thus indicate a major flaw in thedesign analysis. The design goal is 75%, which means that the power that comesout of the injector about equals that on the dump.


Overall Assumptions

In building the free-electron laser, we want to take advantageof existing technology to the greatest extent possible and usedesigns which minimize design and construction time. This willaffect many of our design choices.

Electron beam energy (50 MeV):

The electron beam energy should be as high as is reasonably possiblefor several reasons. First, at higher beam energies the electronbeam has more average power for a given beam current. Since theinjector will limit the beam current to some degree, the maximumbeam power is them limited by the electron beam energy. Second,the emittance should scale as the inverse of the energy and soshould be smaller at higher beam energy. As discussed below,the power output will be limited by the emittance, so it paysto get the emittance as small as possible while maintaining thehighest possible brightness. Thirdly, the higher energy allowsone to use a wiggler with a larger wiggler period. In generala wiggler with a longer period will have a larger gap for a givenmagnetic field. Finally, the higher energy beam is stiffer andless sensitive to stray fields and field jitter.

Weighed against these advantages are the obvious disadvantagesof technical risk, high cost, size, and complexity involved ina higher energy machine. The technical risk is small for a cryomodulewith 40 MV total gradient since this has already been demonstrated. One can therefore get to 50 MeV by using one pass through acryomodule linac after injection at 10 MeV.

Normalized emittance (<13 mm-mrad):

The emittance specifications is set by the shortest wavelengthplanned for the laser (2 µm in this case). The FEL acceptancecondition is that . For a 50 MeV beam anda minimum wavelength of 2 µm the normalized emittance foroptimum operation should be less than 16 mm-mrad. Emittancevalues larger than this will to reduction in both the gain andpower of the laser. The emittance per unit microbunch chargeassumed here is approximately a factor three larger than simulationsof the injector indicate. Initial operation will occur at 3 µmin order to provide a margin of safety on the emittance. Wehave chosen a specification of 13 mm-mrad to unsure third harmoniclasing.

Electron beam average current (5 mA):

As was true with the energy, this should be as high as is reasonablypossible. Since no injector of this type has been operated athigh average current we do not know for sure what the currentlimitation of the injector will be. This is one of the criticaltechnologies in this proposal which must be developed. We havechosen 5 mA as a best guess of what we might be able to achievein this machine. The current may be limited by a number of factorssuch as halo formation, heating of beamline components in thevicinity of the gun or back bombardment of cathode by ions createdby the beam hitting residual gas. The primary limit to beam currentfrom these sources will be cathode lifetime. Boeing has demonstrated35 mA operation from a photocathode gun for periods a few hours. We therefore think that the injector should be able to achieve5 mA operation if only for short periods of time. If the injectorcan produce more than 5 mA of average current, the rf windowsfor the injector cryounit will limit the current to slightly morethan 5 mA.

Electron bunch repetition rate (18.713-74.85 MHz):

The minimum repetition rate of the electron pulses duringoperation is defined by the round trip frequency of the opticalcavity which is given by c/2L.. This frequency must bean even subharmonic of the accelerator frequency and also of thephotocathode drive laser. We presently own a drive laser witha frequency of 74.85 MHz. This frequency is the 20th subharmonicofthe accelerator frequency. This subharmonic is easy to generateby standard frequency division techniques. The cavity lengthmust therefore be a multiple of 2.0028 meters in length. We chooseto operate with an 8 meter to enable cavity alignment withoutactive contols. This means that the minimum repetition rate willbe 18.713 MHz.

The maximum repetition rate is determined by the minimum chargewhich is usable by the laser. The maximum charge at a given frequencyis limited by the 5 mA current limit. Simulations indicate thatthe laser should operate well with a charge varying from 67 pCup to 270 pC. All of these values give a power output of approximately1 kW at 2-4 µm assuming that the longitudinal and transverseemittances scale roughly as the square root of the charge.

Charge per bunch (65-270 pC):

This is determined from the average current and the bunch repetitionrate as noted above. Unlike the UVFEL, the charge is not reallylimited by the HOM power load since the beams only pass throughthe cryomodule twice.

Momentum spread (210 keV):

The rms energy spread in a free-electron laser should be lessthan 1/5N (where N is number of wiggler periods). Since we needat least 40 periods in the wiggler to reduce the exhaust energyspread to reasonable levels the energy spread should be less than0.5% at 42 MeV or 210 keV. Since there is no requirement for laserbandwidth, it should be noted that the energy spread might bedue to an energy slew.

Energy stability (4x10-4 rms):

During turn-on it is necessary to optimize the spontaneous powerspectrum. In order to accomplish this the wavelength jitter mustbe much smaller than the width of the spectrum (1.25%). If therms energy stability os 4x10-4 the wavelengthstability will be 8x10-4 or 6% of the spectralwidth. This will lead to 10% rms power fluctuations in the powerout of the spectrometer.

At high frequencies the energy jitter is also limited due to itseffect on timing jitter induced via the energy jitter interactingwith the M56 of the transport arc. As notedbelow, timing jitter at a high enough frequency will cause largepower fluctuations. The energy jitter condition in this caseis given by:

Note that this applies only to energy jitter which is uncorrelatedwith the timing jitter. Jitter which is correlated with the timingjitter will at least partially cancel the timing jitter.

Peak current (50A) :

The IRFEL gain is approximately linearly proportional the peakcurrent if the energy spread and emittance and are kept constant. Calculations indicate that a peak current of 50A is sufficientto operate in the infrared.

Bunch length (<1 ps rms@135 pC):

This bunch length and the assumed charge give the necessary peakcurrent. For a higher charge the bunch could be longer. For shorterbunch lengths it is desirable to have a shorter length.

Betatron functions (50 cm at wiggler center):

The gain is weakly dependent on the focusing of the electronbeam in the wiggler due to the fact that increasing the focusingincreases the gain due to the improved filling factor and decreasesthe gain due to the increase in the angular spread of the beam. The two effect nearly cancel and cause the gain to be almostindependent of the betatron function in the wiggler. If one wantsthe maximum beam extraction however it is necessary to put asmuch of the electron beam inside the optical mode as possible. One would like the optical mode to be as small as possible tomaximize the mode size at the mirrors. If one applies the constraintthat the electron beam must be smaller than the optical mode andoptimizes the gain versus focusing one finds that the optimumbetatron match is a round beam with a betatron function valueof 50 cm. The matched beta function in the vertical directionis nearly 50 cm so the beam must be matched to a vertical waista the wiggler entrance and a horizontal waist at the wiggler center.This will produce a waist with equal 50 cm beta functions in thewiggler center.

Dispersion in the wigglers (<2 cm):

If we assume that the energy spread is equal to the maximum allowedspread and that the emittance is equal to the maximum specifiedemittance, then we want the beam size increase due to dispersionto be small. If the dispersion is less than 2 cm the gain reductiondue to the increase in the beam size will be less than 10%. Notethat the dispersion can be higher if the energy spread is smallerthan the maximum allowed value.

Timing jitter (<4x10-9/fmrms):

The optical cavity in an RF linac based FEL must haveits round trip travel time precisely matched to the arrival timeof the electron bunches so that the previously emitted bunchesoverlap the fresh electron bunches. Simulations of pulse propagationin FELs indicate that, in order to keep the peak to peak fluctuationssmaller than 10% it is necessary to keep the cavity length stableto less than 0.05GNl. For our casethe gain G is 0.3, the number of wiggler periods N is 40 and theshortest wavelength is 2 µm. One must therefore keep thecavity length constant to 1.2 µm peak to peak. The arrivaltime must be kept constant to the same precision. A change inthe frequency of the pulses is limited by the constraint:

This level of accuracy is already achieved in the CEBAF masteroscillator. At high frequencies however, this implies that therms phase noise must be small and must fall of inversely withthe frequency to keep the frequency modulation less than thisvalue. From the frequency modulation constraint it is possibleto derive a timing jitter constraint of 4x10-9/fm whereit has been assumed that the peak to peak jitter is six timesthe rms jitter and fm is the frequencyof the jitter.

Small signal gain to loss ratio (>3):

If the ratio of small signal gain to saturated gain (the saturatedgain is equal to the cavity losses) is greater than three, onefinds that the laser is fairly insensitive to fluctuations inthe gain. This is because the slope of the gain versus intensityis very steep when the gain has been reduced by this amount bysaturation. For lower gain to loss ratio we will not be ableto keep the laser power fluctuations smaller than 10% due to avery high sensitivity of the saturation point on the electronbeam parameters.

Current jitter (<2% peak-to-)peak):

A 2% current jitter will create a 2% jitter in the gain. Sincewe will be in saturation this will not strongly affect the extractionefficiency. This means that the laser power will also fluctuateby approximately 2% peak to peak. As noted above the frequencyjitter caused by phase noise on the electron beam will dominatethe laser power jitter. If the current jitter is as small as2% it should not appreciably increase the jitter in the laserpower.

Output coupling efficiency (>85%):

If we want 1 kW of laser power and have 210 kW of circulatingbeam (90% of which is usable for lasing), and the efficiency ofthe laser is 1/4N, where N is number of wiggler periods, thenwe extract 1180 W from the electron beam and need 85% output couplingefficiency from the optical cavity.

Beam angle jitter ( 250 µrad rms):

The beam angle jitter should be small compared to the opticalmode divergence, which is 1.25 mrad. If the beam angle jitteris less than 250 µrad rms, the optical mode quality willnot be appreciably effected.

Beam position (100 µm rms):

The beam must be centered in the optical mode to withina small fraction of the beam waist of the optical mode. The opticalmode waist is 500 µm at 2 µm. If the beam is stableto 100 µm in transverse position, the laser gain should bereasonably stable.

Maintained by